1 NVIDIA Accelerated FreeBSD Graphics Driver README and Installation Guide
4 Last Updated: Tue Feb 3 09:57:25 PST 2009
5 Most Recent Driver Version: 180.29
9 2701 San Tomas Expressway
16 ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS,
17 DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, "MATERIALS")
18 ARE BEING PROVIDED "AS IS." NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED,
19 STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS
20 ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A
21 PARTICULAR PURPOSE. Information furnished is believed to be accurate and
22 reliable. However, NVIDIA Corporation assumes no responsibility for the
23 consequences of use of such information or for any infringement of patents or
24 other rights of third parties that may result from its use. No license is
25 granted by implication or otherwise under any patent or patent rights of
26 NVIDIA Corporation. Specifications mentioned in this publication are subject
27 to change without notice. This publication supersedes and replaces all
28 information previously supplied. NVIDIA Corporation products are not
29 authorized for use as critical components in life support devices or systems
30 without express written approval of NVIDIA Corporation.
32 NVIDIA, the NVIDIA logo, NVIDIA nForce, GeForce, NVIDIA Quadro, Vanta, TNT2,
33 TNT, RIVA, RIVA TNT, Quincunx Antialiasing, and TwinView are registered
34 trademarks or trademarks of NVIDIA Corporation in the United States and/or
37 FreeBSD is a registered trademark of the FreeBSD Foundation. Linux is a
38 registered trademark of Linus Torvalds. Intel and Pentium are registered
39 trademarks of Intel Corporation. Athlon is a registered trademark of Advanced
40 Micro Devices. OpenGL is a registered trademark of Silicon Graphics Inc. PCI
41 Express is a registered trademark and/or service mark of PCI-SIG. Windows is a
42 registered trademark of Microsoft Corporation in the United States and other
43 countries. Other company and product names may be trademarks or registered
44 trademarks of the respective owners with which they are associated.
47 Copyright 2006 - 2008 NVIDIA Corporation. All rights reserved.
49 ______________________________________________________________________________
52 ______________________________________________________________________________
54 Chapter 1. Introduction
55 Chapter 2. Installing the NVIDIA Driver
56 Chapter 3. Using Linux Compatibility Support
57 Chapter 4. Configuring X for the NVIDIA Driver
58 Chapter 5. Frequently Asked Questions
59 Chapter 6. Common Problems
60 Chapter 7. Known Issues
61 Chapter 8. Specifying OpenGL Environment Variable Settings
62 Chapter 9. Configuring AGP
63 Chapter 10. Configuring TwinView
64 Chapter 11. Configuring GLX in Xinerama
65 Chapter 12. Configuring Multiple X Screens on One Card
66 Chapter 13. Configuring TV-Out
67 Chapter 14. Using the XRandR Extension
68 Chapter 15. Configuring a Notebook
69 Chapter 16. Programming Modes
70 Chapter 17. Configuring Flipping and UBB
71 Chapter 18. Using the X Composite Extension
72 Chapter 19. Using the nvidia-settings Utility
73 Chapter 20. Configuring SLI and Multi-GPU FrameRendering
74 Chapter 21. Configuring Frame Lock and Genlock
75 Chapter 22. Configuring SDI Video Output
76 Chapter 23. Configuring Depth 30 Displays
77 Chapter 24. NVIDIA Contact Info and Additional Resources
79 Chapter 26. Acknowledgements
81 Appendix A. Minimum Software Requirements
82 Appendix B. Installed Components
83 Appendix C. The Sysctl Interface
84 Appendix D. Configuring Low-level Parameters
85 Appendix A. Supported NVIDIA GPU Products
86 Appendix B. X Config Options
87 Appendix C. Display Device Names
88 Appendix D. GLX Support
89 Appendix E. Dots Per Inch
90 Appendix F. XvMC Support
91 Appendix G. VDPAU Support
92 Appendix H. Tips for New FreeBSD Users
94 ______________________________________________________________________________
96 Chapter 1. Introduction
97 ______________________________________________________________________________
100 1A. ABOUT THE NVIDIA ACCELERATED FREEBSD GRAPHICS DRIVER
102 The NVIDIA Accelerated FreeBSD Graphics Driver brings accelerated 2D
103 functionality and high-performance OpenGL support to FreeBSD x86 with the use
104 of NVIDIA graphics processing units (GPUs).
106 These drivers provide optimized hardware acceleration for OpenGL and X
107 applications and support nearly all recent NVIDIA GPU products (see Appendix E
108 for a complete list of supported GPUs). TwinView, TV-Out and flat panel
109 displays are also supported.
112 1B. ABOUT THIS DOCUMENT
114 This document provides instructions for the installation and use of the NVIDIA
115 Accelerated FreeBSD Graphics Driver. Chapter 2, Chapter 3 and Chapter 4 walk
116 the user through the process of downloading, installing and configuring the
117 driver. Chapter 5 addresses frequently asked questions about the installation
118 process, and Chapter 6 provides solutions to common problems. The remaining
119 chapters include details on different features of the NVIDIA FreeBSD Driver.
120 Frequently asked questions about specific tasks are included in the relevant
124 1C. ABOUT THE AUDIENCE
126 It is assumed that the user and reader of this document has at least a basic
127 understanding of FreeBSD techniques and terminology. However, new FreeBSD
128 users can refer to Appendix L for details on parts of the installation
132 1D. ADDITIONAL INFORMATION
134 In case additional information is required, Chapter 24 provides contact
135 information for NVIDIA FreeBSD driver resources, as well as a brief listing of
138 ______________________________________________________________________________
140 Appendix A. Minimum Software Requirements
141 ______________________________________________________________________________
143 The official minimum software requirements for the NVIDIA FreeBSD Graphics
144 Driver are as follows:
146 Software Element Min Requirement
147 ---------------------------------- ----------------------------------
148 Kernel FreeBSD 5-STABLE (5.3 or later)
149 XFree86/X.Org 4.2/6.7.0
151 Additionally, the kernel source tree must be installed in /usr/src/sys
152 (package 'ssys' installed)
154 Note that FreeBSD -STABLE versions older than FreeBSD 5.3 and FreeBSD 6.x/7.x
155 -CURRENT development snapshots are not supported.
157 ______________________________________________________________________________
159 Chapter 2. Installing the NVIDIA Driver
160 ______________________________________________________________________________
162 This installation procedure will likely be simplified further in the future,
163 but for the moment you will need to download the NVIDIA FreeBSD Graphics
164 Driver archives from the NVIDIA website, extract them to a temporary location
165 of your choice, and run the following from the root of the extracted directory
170 This will compile the NVIDIA FreeBSD kernel module, install it, and kldload
171 it. It will also remove any conflicting OpenGL libraries, and install the
172 NVIDIA OpenGL libraries. The '/dev/nvidia' device files will be created
173 (unless the system is using devfs), and your '/boot/loader.conf' file will be
174 updated to automatically load the NVIDIA kernel module on boot, as well as the
175 Linux ABI compatibility module should you not have it compiled into your
178 ______________________________________________________________________________
180 Appendix B. Installed Components
181 ______________________________________________________________________________
183 The NVIDIA Accelerated FreeBSD Graphics Driver consists of the following
186 Installed File Location
187 ---------------------------------- ----------------------------------
188 nvidia.ko /boot/modules
189 libGL.so /usr/lib/xorg
190 libGL.so.1 /usr/lib/xorg
191 libnvidia-tls.so /usr/lib/xorg
192 libnvidia-tls.so.1 /usr/lib/xorg
193 libnvidia-cfg.so /usr/lib/xorg
194 libnvidia-cfg.so.1 /usr/lib/xorg
195 libGLcore.so /usr/lib/xorg
196 libGLcore.so.1 /usr/lib/xorg
197 nvidia_drv.so /usr/lib/xorg/modules/drivers
198 libnvidia-wfb.so (optional) /usr/lib/xorg/modules
199 libwfb.so /usr/lib/xorg/modules/drivers
200 libglx.so /usr/lib/xorg/modules/extensions
201 libglx.so.1 /usr/lib/xorg/modules/extensions
202 libvdpau.so /usr/lib/xorg
203 libvdpau.so.1 /usr/lib/xorg
204 libvdpau_trace.so /usr/lib/xorg
205 libvdpau_trace.so.1 /usr/lib/xorg
206 libvdpau_nvidia.so /usr/lib/xorg
207 libvdpau_nvidia.so.1 /usr/lib/xorg
208 nvidia-xconfig /usr/bin
209 nvidia-xconfig.1 /usr/man/man1
210 nvidia-settings /usr/bin
211 nvidia-settings.1 /usr/man/man1
217 libGL.so.180.29 /compat/linux/usr/lib
218 libnvidia-tls.so.180.29 /compat/linux/usr/lib
219 libGLcore.so.180.29 /compat/linux/usr/lib
220 libvdpau.so.180.29 /compat/linux/usr/lib
221 libvdpau_trace.so.180.29 /compat/linux/usr/lib
222 libvdpau_nvidia.so.180.29 /compat/linux/usr/lib
225 ______________________________________________________________________________
227 Chapter 3. Using Linux Compatibility Support
228 ______________________________________________________________________________
230 If you wish to run Linux OpenGL applications on your FreeBSD computer, you
231 will need to make sure that several prerequisites are met.
233 First, you should follow the basic Linux compatibility installation guide in
234 the FreeBSD Handbook (install the linux_base package, etc). Once the basic
235 components are in place, you will need to install the NVIDIA Linux OpenGL
236 libraries in '/compat/linux/usr/lib' (do not brandelf them!); if the
237 '/compat/linux/usr/lib/' directory exists when you install the FreeBSD driver,
238 the Linux compatibility OpenGL libraries will automatically be installed.
240 Additionally, the 'nvidia.ko' kernel module needs to be built with support for
241 the Linux ABI compatibility layer. This is the case by default; as a
242 consequence, the 'nvidia.ko' kernel module requires the 'linux.ko' module to
245 Note: If you have no need for Linux ABI compatibility and do not wish to load
246 'linux.ko', you can build the 'nvidia.ko' kernel module without support for
247 the Linux ABI compatibility layer (see 'nv-freebsd.h' for details).
249 ______________________________________________________________________________
251 Chapter 4. Configuring X for the NVIDIA Driver
252 ______________________________________________________________________________
254 The X configuration file provides a means to configure the X server. This
255 section describes the settings necessary to enable the NVIDIA driver. A
256 comprehensive list of parameters is provided in Appendix F.
258 The NVIDIA Driver includes a utility called nvidia-xconfig, which is designed
259 to make editing the X configuration file easy. You can also edit it by hand.
262 4A. USING NVIDIA-XCONFIG TO CONFIGURE THE X SERVER
264 nvidia-xconfig will find the X configuration file and modify it to use the
265 NVIDIA X driver. In most cases, you can simply answer "Yes" when the installer
266 asks if it should run it. If you need to reconfigure your X server later, you
267 can run nvidia-xconfig again from a terminal. nvidia-xconfig will make a
268 backup copy of your configuration file before modifying it.
270 Note that the X server must be restarted for any changes to its configuration
273 More information about nvidia-xconfig can be found in the nvidia-xconfig
274 manual page by running.
281 4B. MANUALLY EDITING THE CONFIGURATION FILE
283 In April 2004 the X.Org Foundation released an X server based on the XFree86
284 server. While your release may use the X.Org X server, rather than XFree86,
285 the differences between the two should have no impact on NVIDIA FreeBSD users
288 o The X.Org configuration file is '/etc/X11/xorg.conf' while the XFree86
289 configuration file is '/etc/X11/XF86Config'. The files use the same
290 syntax. This document refers to both files as "the X config file".
292 o The X.Org log file is '/var/log/Xorg.#.log' while the XFree86 log file is
293 '/var/log/XFree86.#.log' (where '#' is the server number -- usually 0).
294 The format of the log files is nearly identical. This document refers to
295 both files as "the X log file".
297 In order for any changes to be read into the X server, you must edit the file
298 used by the server. While it is not unreasonable to simply edit both files, it
299 is easy to determine the correct file by searching for the line
301 (==) Using config file:
303 in the X log file. This line indicates the name of the X config file in use.
305 If you do not have a working X config file, there are a few different ways to
306 obtain one. A sample config file is included both with the XFree86
307 distribution and with the NVIDIA driver package (at
308 '/usr/X11R6/share/doc/NVIDIA_GLX-1.0/'). The 'nvidia-xconfig' utility,
309 provided with the NVIDIA driver package, can generate a new X configuration
310 file. Additional information on the X config syntax can be found in the
311 XF86Config manual page (`man XF86Config` or `man xorg.conf`).
313 If you have a working X config file for a different driver (such as the "nv"
314 or "vesa" driver), then simply edit the file as follows.
322 and replace it with the line:
326 Remove the following lines:
331 In the "Module" section of the file, add the line (if it does not already
336 If the X config file does not have a "Module" section, you can safely skip the
337 last step if the X server installed on your system is an X.Org X server or an
338 XFree86 X release version 4.4.0 or greater. If you are using an older XFree86
339 X server, add the following to your X config file:
349 There are numerous options that may be added to the X config file to tune the
350 NVIDIA X driver. See Appendix F for a complete list of these options.
352 Once you have completed these edits to the X config file, you may restart X
353 and begin using the accelerated OpenGL libraries. After restarting X, any
354 OpenGL application should automatically use the new NVIDIA libraries. (NOTE:
355 If you encounter any problems, see Chapter 6 for common problem diagnoses.)
357 ______________________________________________________________________________
359 Chapter 5. Frequently Asked Questions
360 ______________________________________________________________________________
362 This section provides answers to frequently asked questions associated with
363 the NVIDIA FreeBSD x86 Driver and its installation. Common problem diagnoses
364 can be found in Chapter 6 and tips for new users can be found in Appendix L.
365 Also, detailed information for specific setups is provided in the Appendices.
370 Q. Where should I start when diagnosing display problems?
372 A. One of the most useful tools for diagnosing problems is the X log file in
373 '/var/log'. Lines that begin with "(II)" are information, "(WW)" are
374 warnings, and "(EE)" are errors. You should make sure that the correct
375 config file (i.e. the config file you are editing) is being used; look for
376 the line that begins with:
378 (==) Using config file:
380 Also make sure that the NVIDIA driver is being used, rather than the "nv"
381 or "vesa" driver. Search for
383 (II) LoadModule: "nvidia"
385 Lines from the driver should begin with:
391 Q. How can I increase the amount of data printed in the X log file?
393 A. By default, the NVIDIA X driver prints relatively few messages to stderr
394 and the X log file. If you need to troubleshoot, then it may be helpful to
395 enable more verbose output by using the X command line options -verbose and
396 -logverbose, which can be used to set the verbosity level for the 'stderr'
397 and log file messages, respectively. The NVIDIA X driver will output more
398 messages when the verbosity level is at or above 5 (X defaults to verbosity
399 level 1 for 'stderr' and level 3 for the log file). So, to enable verbose
400 messaging from the NVIDIA X driver to both the log file and 'stderr', you
401 could start X with the verbosity level set to 5, by doing the following
403 % startx -- -verbose 5 -logverbose 5
407 Q. I have read that the NVIDIA FreeBSD Driver is not a native driver, but sits
408 on top of the Linux ABI compatibility layer. Is this true?
410 A. No, the NVIDIA FreeBSD Graphics Driver is a native driver. It does provide
411 Linux OpenGL libraries in addition to the native, FreeBSD libraries to
412 enable users to run Linux OpenGL applications.
415 Q. Is the NVIDIA FreeBSD Accelerated Graphics Driver thread-safe?
417 A. This release is thread-safe on FreeBSD 5.3 or later systems making use of
418 the libpthread or libthr KSE threading libraries. On these systems, the
419 NVIDIA Linux ABI compatibility libraries are fully thread-safe as well.
422 Q. Why can't the Linux compatibility libraries correctly determine if they are
423 used in a multithreaded application?
425 A. The Linux compatibility libraries are not able to correctly determine if
426 they are used in a multithreaded application because the %gs segment
427 register is not initialized correctly for Linux compatibility.
429 The '__GL_SINGLE_THREADED' environment variable (set to "1") can be used to
430 work around this issue, but at the cost of thread-safeness.
433 Q. Why do applications that use DGA graphics fail?
435 A. The NVIDIA driver does not support the graphics component of the
436 XFree86-DGA (Direct Graphics Access) extension. Applications can use the
437 XDGASelectInput() function to acquire relative pointer motion, but
438 graphics-related functions such as XDGASetMode() and XDGAOpenFramebuffer()
441 The graphics component of XFree86-DGA is not supported because it requires
442 a CPU mapping of framebuffer memory. As graphics cards ship with increasing
443 quantities of video memory, the NVIDIA X driver has had to switch to a more
444 dynamic memory mapping scheme that is incompatible with DGA. Furthermore,
445 DGA does not cooperate with other graphics rendering libraries such as Xlib
446 and OpenGL because it accesses GPU resources directly.
448 NVIDIA recommends that applications use OpenGL or Xlib, rather than DGA,
449 for graphics rendering. Using rendering libraries other than DGA will yield
450 better performance and improve interoperability with other X applications.
453 Q. My kernel log contains messages that are prefixed with "Xid"; what do these
456 A. "Xid" messages indicate that a general GPU error occurred, most often due
457 to the driver misprogramming the GPU or to corruption of the commands sent
458 to the GPU. These messages provide diagnostic information that can be used
459 by NVIDIA to aid in debugging reported problems.
462 Q. On what NVIDIA hardware is the EXT_framebuffer_object OpenGL extension
465 A. EXT_framebuffer_object is supported on GeForce FX, Quadro FX, and newer
469 Q. I use the Coolbits overclocking interface to adjust my graphics card's
470 clock frequencies, but the defaults are reset whenever X is restarted. How
471 do I make my changes persistent?
473 A. Clock frequency settings are not saved/restored automatically by default to
474 avoid potential stability and other problems that may be encountered if the
475 chosen frequency settings differ from the defaults qualified by the
476 manufacturer. You can use the command line below in '~/.xinitrc' to
477 automatically apply custom clock frequency settings when the X server is
480 # nvidia-settings -a GPUOverclockingState=1 -a
481 GPU2DClockFreqs=<GPU>,<MEM> -a GPU3DClockFreqs=<GPU>,<MEM>
483 Here '<GPU>' and '<MEM>' are the desired GPU and video memory frequencies
484 (in MHz), respectively.
487 Q. Why is the refresh rate not reported correctly by utilities that use the
488 XRandR X extension (e.g., the GNOME "Screen Resolution Preferences" panel,
491 A. The XRandR X extension is not presently aware of multiple display devices
492 on a single X screen; it only sees the MetaMode bounding box, which may
493 contain one or more actual modes. This means that if multiple MetaModes
494 have the same bounding box, XRandR will not be able to distinguish between
497 In order to support DynamicTwinView, the NVIDIA X driver must make each
498 MetaMode appear to be unique to XRandR. Presently, the NVIDIA X driver
499 accomplishes this by using the refresh rate as a unique identifier.
501 You can use `nvidia-settings -q RefreshRate` to query the actual refresh
502 rate on each display device.
504 This behavior can be disabled by setting the X configuration option
505 "DynamicTwinView" to FALSE.
507 For details, see Chapter 10.
510 Q. Why does starting certain applications result in Xlib error messages
511 indicating extensions like "XFree86-VidModeExtension" or "SHAPE" are
514 A. If your X config file has a "Module" section that does not list the
515 "extmod" module, some X server extensions may be missing, resulting in
516 error messages of the form:
518 Xlib: extension "SHAPE" missing on display ":0.0"
519 Xlib: extension "XFree86-VidModeExtension" missing on display ":0.0"
520 Xlib: extension "XFree86-DGA" missing on display ":0.0"
522 You can solve this problem by adding the line below to your X config file's
529 ______________________________________________________________________________
531 Chapter 6. Common Problems
532 ______________________________________________________________________________
534 This section provides solutions to common problems associated with the NVIDIA
537 Q. My X server fails to start, and my X log file contains the error:
539 (EE) NVIDIA(0): The NVIDIA kernel module does not appear to
540 (EE) NVIDIA(0): be receiving interrupts generated by the NVIDIA
542 (EE) NVIDIA(0): device PCI:x:x:x. Please see the COMMON PROBLEMS
543 (EE) NVIDIA(0): section in the README for additional information.
546 A. This can be caused by a variety of problems, such as PCI IRQ routing
547 errors, I/O APIC problems or conflicts with other devices sharing the IRQ
550 If possible, configure your system such that your graphics card does not
551 share its IRQ with other devices (try moving the graphics card to another
552 slot if applicable, unload/disable the driver(s) for the device(s) sharing
553 the card's IRQ, or remove/disable the device(s)).
556 Q. My X server fails to start, and my X log file contains the error:
558 (EE) NVIDIA(0): The interrupt for NVIDIA graphics device PCI:x:x:x
559 (EE) NVIDIA(0): appears to be edge-triggered. Please see the COMMON
560 (EE) NVIDIA(0): PROBLEMS section in the README for additional
564 A. An edge-triggered interrupt means that the kernel has programmed the
565 interrupt as edge-triggered rather than level-triggered in the Advanced
566 Programmable Interrupt Controller (APIC). Edge-triggered interrupts are not
567 intended to be used for sharing an interrupt line between multiple devices;
568 level-triggered interrupts are the intended trigger for such usage. When
569 using edge-triggered interrupts, it is common for device drivers using that
570 interrupt line to stop receiving interrupts. This would appear to the end
571 user as those devices no longer working, and potentially as a full system
572 hang. These problems tend to be more common when multiple devices are
573 sharing that interrupt line.
576 Q. X starts for me, but OpenGL applications terminate immediately.
578 A. If X starts but you have trouble with OpenGL, you most likely have a
579 problem with other libraries in the way, or there are stale symlinks. See
580 Appendix B for details.
582 You should also check that the correct extensions are present;
586 should show the "GLX" and "NV-GLX" extensions present. If these two
587 extensions are not present, then there is most likely a problem loading the
588 glx module, or it is unable to implicitly load GLcore. Check your X config
589 file and make sure that you are loading glx (see Chapter 4). If your X
590 config file is correct, then check the X log file for warnings/errors
591 pertaining to GLX. Also check that all of the necessary symlinks are in
592 place (refer to Appendix B).
595 Q. When Xinerama is enabled, my stereo glasses are shuttering only when the
596 stereo application is displayed on one specific X screen. When the
597 application is displayed on the other X screens, the stereo glasses stop
600 A. This problem occurs with DDC and "blue line" stereo glasses, that get the
601 stereo signal from one video port of the graphics card. When a X screen
602 does not display any stereo drawable the stereo signal is disabled on the
603 associated video port.
605 Forcing stereo flipping allows the stereo glasses to shutter continuously.
606 This can be done by enabling the OpenGL control "Force Stereo Flipping" in
607 nvidia-settings, or by setting the X configuration option
608 "ForceStereoFlipping" to "1".
611 Q. Stereo is not in sync across multiple displays.
613 A. There are two cases where this may occur. If the displays are attached to
614 the same GPU, and one of them is out of sync with the stereo glasses, you
615 will need to reconfigure your monitors to drive identical mode timings; see
616 Chapter 16 for details.
618 If the displays are attached to different GPUs, the only way to synchronize
619 stereo across the displays is with a G-Sync device, which is only supported
620 by certain Quadro cards. See Chapter 21 for details. This applies to
621 seperate GPUs on seperate cards as well as seperate GPUs on the same card,
622 such as Quadro FX 4500 X2. Note that the Quadro FX 4500 X2 only provides a
623 single DIN connector for stereo, tied to the bottommost GPU. In order to
624 synchronize onboard stereo on the other GPU you must use a G-Sync device.
627 Q. X fails to start, and during boot up time I get error messages
629 nvidia0: NVRM: NVIDIA REG resource alloc failed.
633 nvidia0: NVRM: NVIDIA IRQ resource alloc failed.
636 A. The system BIOS has not properly set up your graphics card; FreeBSD can't
637 currently set up PCI devices that the BIOS leaves unconfigured. Uncheck
638 "PNP-OS" in your system BIOS.
641 Q. X fails to start, and during boot up time I get the following error
644 nvidia0: NVRM: NVIDIA MEM resource alloc failed.
647 A. On certain FreeBSD kernels, it may be necessary to add the following line
648 to '/boot/loader.conf':
650 hw.pci.allow_unsupported_io_range="1"
652 This should allow the NVIDIA kernel module to attach.
655 Q. Some OpenGL applications fail to start, and my X log file contains error
656 messages of the form:
658 (EE) NVIDIA(0): Failed to obtain a shared memory identifier...
662 (EE) NVIDIA(0): Failed to attach to shared memory segment...
665 A. The NVIDIA driver may require more IPC resources than are allocated by
666 default. If this happens, you may be able to work around this problem by
667 increasing your system's resource limits by editing the file
668 '/boot/loader.conf' and adding:
674 The values above were chosen conservatively, you may need to tweak them to
675 meet your requirements.
678 Q. My X server fails to start, and my X log file contains the error:
680 (EE) NVIDIA(0): Failed to initialize the NVIDIA kernel module!
683 A. Nothing will work if the NVIDIA kernel module does not function properly.
684 If you see anything in the X log file like
686 (EE) NVIDIA(0): Failed to initialize the NVIDIA kernel module!
688 then there is most likely a problem with the NVIDIA kernel module.
690 The NVIDIA kernel module may print error messages indicating a problem --
691 to view these messages check the output of `dmesg`, '/var/log/messages', or
692 wherever syslog is directed to place kernel messages. These messages are
693 prepended with "NVRM".
696 Q. When I attempt to start `nvidia-settings`, I get an error message of the
699 Shared object "libgtk-x11-2.0.so.400" not found, required by
703 A. Due to differences between the gtk+-2.x ports packages included with
704 different FreeBSD 5.x releases, the prebuilt nvidia-settings binary shipped
705 with the NVIDIA driver may not work with FreeBSD releases more recent than
708 If you have a recent ports package of gtk+-2.x and gmake installed on your
709 system, you can build the nvidia-installer utility from source to solve
712 Download nvidia-settings-1.0.tar.gz (or the latest version) from
713 ftp://download.nvidia.com/XFree86/nvidia-settings You can then extract,
714 build and install it (to '/usr/local/bin') with:
720 Q. When I attempt to run `nvidia-xconfig` after the NVIDIA FreeBSD graphics
721 driver installation, I get an error message of the form:
723 nvidia-xconfig: Command not found.
726 A. Depending on the shell you are using, you may need to force it to recompute
727 its internal table of executable files present in the directories listed in
728 the '$PATH' variable. Assuming you are using the FreeBSD default shell you
729 can do so by issuing the command:
735 Q. When I attempt to start a Linux application as 'root', I get the error
738 NVIDIA: failed to execute '/sbin/modprobe': No such file or directory.
741 A. When initialized by an application executed with 'root' privileges, the
742 NVIDIA Linux OpenGL library, shipped with the NVIDIA FreeBSD graphics
743 driver for Linux ABI compatibility, will attempt to load the NVIDIA Linux
744 kernel module and fail because /sbin/modprobe is absent. You can work
745 around this problem by creating a symbolic link from '/usr/bin/true' to
746 '/compat/linux/sbin/modprobe':
748 % ln -s /usr/bin/true /compat/linux/sbin/modprobe
752 Q. My system runs, but seems unstable.
754 A. Your stability problems may be AGP-related. See Chapter 9 for details.
757 Q. OpenGL applications are running slowly
759 A. The application is probably using a different library that still remains on
760 your system, rather than the NVIDIA supplied OpenGL library. See Appendix B
764 Q. There are problems running Quake2.
766 A. Quake2 requires some minor setup to get it going. First, in the Quake2
767 directory, the install creates a symlink called 'libGL.so' that points at
768 'libMesaGL.so'. This symlink should be removed or renamed. Second, in order
769 to run Quake2 in OpenGL mode, you must type
771 % quake2 +set vid_ref glx +set gl_driver libGL.so
773 Quake2 does not seem to support any kind of full-screen mode, but you can
774 run your X server at the same resolution as Quake2 to emulate full-screen
778 Q. I am using either nForce of nForce2 internal graphics, and I see warnings
779 like this in my X log file:
781 Not using mode "1600x1200" (exceeds valid memory bandwidth usage)
784 A. Integrated graphics have more strict memory bandwidth limitations that
785 limit the resolution and refresh rate of the modes you request. To work
786 around this, you can reduce the maximum refresh rate by lowering the upper
787 value of the VertRefresh range in the 'Monitor' section of your X config
788 file. Though not recommended, you can disable the memory bandwidth test
789 with the NoBandWidthTest X config file option.
792 Q. X takes a long time to start (possibly several minutes).
794 A. Most of the X startup delay problems we have found are caused by incorrect
795 data in video BIOSes about what display devices are possibly connected or
796 what i2c port should be used for detection. You can work around these
797 problems with the X config option IgnoreDisplayDevices (see the description
801 Q. Fonts are incorrectly sized after installing the NVIDIA driver.
803 A. Incorrectly sized fonts are generally caused by incorrect DPI (Dots Per
804 Inch) information. You can check what X thinks the physical size of your
805 monitor is, by running:
807 % xdpyinfo | grep dimensions
809 This will report the size in pixels, and in millimeters.
811 If these numbers are wrong, you can correct them by modifying the X
812 server's DPI setting. See Appendix I for details.
815 Q. General problems with ALi chipsets
817 A. There are some known timing and signal integrity issues on ALi chipsets.
818 The following tips may help stabilize problematic ALI systems:
820 o Disable TURBO AGP MODE in the BIOS.
822 o When using a P5A upgrade to BIOS Revision 1002 BETA 2.
824 o When using 1007, 1007A or 1009 adjust the IO Recovery Time to 4
827 o AGP is disabled by default on some ALi chipsets (ALi1541, ALi1647) to
828 work around severe system stability problems with these chipsets. See
829 the comments for EnableALiAGP in 'nv-reg.h' to force AGP on anyway.
833 Q. Using GNOME configuration utilities, I am unable to get a resolution above
836 A. The installation of GNOME provided in operating systems such as Red Hat
837 Enterprise Linux 4 and Solaris 10 Update 2 contain several competing
838 interfaces for specifying resolution:
841 'System Settings' -> 'Display'
844 which will update the X configuration file, and
847 'Applications' -> 'Preferences' -> 'Screen Resolution'
850 which will update the per-user screen resolution using the XRandR
851 extension. Your desktop resolution will be limited to the smaller of the
852 two settings. Be sure to check the setting of each.
855 Q. X does not restore the VGA console when run on a TV. I get this error
856 message in my X log file:
858 Unable to initialize the X int10 module; the console may not be
859 restored correctly on your TV.
862 A. The NVIDIA X driver uses the X Int10 module to save and restore console
863 state on TV out, and will not be able to restore the console correctly if
864 it cannot use the Int10 module. If you have built the X server yourself,
865 please be sure you have built the Int10 module. If you are using a build of
866 the X server provided by your operating system and are missing the Int10
867 module, contact your operating system distributor.
870 Q. OpenGL applications don't work, and my X log file contains the error:
872 (EE) NVIDIA(0): Unable to map device node /dev/zero with read, write, and
873 (EE) NVIDIA(0): execute privileges. The GLX extension will be disabled
874 (EE) NVIDIA(0): on this X screen. Please see the COMMON PROBLEMS
875 (EE) NVIDIA(0): section in the README for more information.
878 A. The NVIDIA OpenGL driver must be able to map the '/dev/zero' device node
879 with read, write, and execute privileges in order to function correctly.
880 The driver needs this ability to allocate executable memory, which is used
881 for optimizations that require generating code at run-time. Currently, GLX
882 cannot run without these optimizations.
884 Check that your '/dev' filesystem is set up correctly. In particular,
885 mounting the '/dev' file system with the 'noexec' option will cause this to
886 happen. If you haven't changed the configuration of your '/dev' filesystem,
887 please contact your operating system distributor.
890 ______________________________________________________________________________
892 Chapter 7. Known Issues
893 ______________________________________________________________________________
895 The following problems still exist in this release and are in the process of
902 If you are using a notebook see the "Known Notebook Issues" in Chapter 15.
906 When FSAA is enabled (the __GL_FSAA_MODE environment variable is set to a
907 value that enables FSAA and a multisample visual is chosen), the rendering
908 may be corrupted when resizing the window.
910 libGL DSO finalizer and pthreads
912 When a multithreaded OpenGL application exits, it is possible for libGL's
913 DSO finalizer (also known as the destructor, or "_fini") to be called
914 while other threads are executing OpenGL code. The finalizer needs to free
915 resources allocated by libGL. This can cause problems for threads that are
916 still using these resources. Setting the environment variable
917 "__GL_NO_DSO_FINALIZER" to "1" will work around this problem by forcing
918 libGL's finalizer to leave its resources in place. These resources will
919 still be reclaimed by the operating system when the process exits. Note
920 that the finalizer is also executed as part of dlclose(3), so if you have
921 an application that dlopens(3) and dlcloses(3) libGL repeatedly,
922 "__GL_NO_DSO_FINALIZER" will cause libGL to leak resources until the
923 process exits. Using this option can improve stability in some
924 multithreaded applications, including Java3D applications.
926 XVideo and the Composite X extension
928 XVideo will not work correctly when Composite is enabled unless using
929 X.Org 7.1 or later. See Chapter 18.
931 GLX visuals in Xinerama
933 X servers prior to version 1.5.0 have a limitation in the number of
934 visuals that can be available when Xinerama is enabled. Specifically,
935 visuals with ID values over 255 will cause the server to corrupt memory,
936 leading to incorrect behavior or crashes. In some configurations where
937 many GLX features are enabled at once, the number of GLX visuals will
938 exceed this limit. To avoid a crash, the NVIDIA X driver will discard
939 visuals above the limit. To see which visuals are being discarded, run the
940 X server with the -logverbose 6 option and then check the X server log
943 This section describes problems that will not be fixed. Usually, the source of
944 the problem is beyond the control of NVIDIA. Following is the list of
947 Problems that Will Not Be Fixed
949 Gigabyte GA-6BX Motherboard
951 This motherboard uses a LinFinity regulator on the 3.3 V rail that is only
952 rated to 5 A -- less than the AGP specification, which requires 6 A. When
953 diagnostics or applications are running, the temperature of the regulator
954 rises, causing the voltage to the NVIDIA GPU to drop as low as 2.2 V.
955 Under these circumstances, the regulator cannot supply the current on the
956 3.3 V rail that the NVIDIA GPU requires.
958 This problem does not occur when the graphics card has a switching
959 regulator or when an external power supply is connected to the 3.3 V rail.
961 VIA KX133 and 694X Chip sets with AGP 2x
963 On Athlon motherboards with the VIA KX133 or 694X chip set, such as the
964 ASUS K7V motherboard, NVIDIA drivers default to AGP 2x mode to work around
965 insufficient drive strength on one of the signals.
967 Irongate Chip sets with AGP 1x
969 AGP 1x transfers are used on Athlon motherboards with the Irongate chipset
970 to work around a problem with signal integrity.
972 ALi chipsets, ALi1541 and ALi1647
974 On ALi1541 and ALi1647 chipsets, NVIDIA drivers disable AGP to work around
975 timing issues and signal integrity issues. See Chapter 6 for more
976 information on ALi chipsets.
978 NV-CONTROL versions 1.8 and 1.9
980 Version 1.8 of the NV-CONTROL X Extension introduced target types for
981 setting and querying attributes as well as receiving event notification on
982 targets. Targets are objects like X Screens, GPUs and G-Sync devices.
983 Previously, all attributes were described relative to an X Screen. These
984 new bits of information (target type and target id) were packed in a
985 non-compatible way in the protocol stream such that addressing X Screen 1
986 or higher would generate an X protocol error when mixing NV-CONTROL client
989 This packing problem has been fixed in the NV-CONTROL 1.10 protocol,
990 making it possible for the older (1.7 and prior) clients to communicate
991 with NV-CONTROL 1.10 servers. Furthermore, the NV-CONTROL 1.10 client
992 library has been updated to accommodate the target protocol packing bug
993 when communicating with a 1.8 or 1.9 NV-CONTROL server. This means that
994 the NV-CONTROL 1.10 client library should be able to communicate with any
995 version of the NV-CONTROL server.
997 NVIDIA recommends that NV-CONTROL client applications relink with version
998 1.10 or later of the NV-CONTROL client library (libXNVCtrl.a, in the
999 nvidia-settings-1.0.tar.gz tarball). The version of the client library can
1000 be determined by checking the NV_CONTROL_MAJOR and NV_CONTROL_MINOR
1001 definitions in the accompanying nv_control.h.
1003 The only web released NVIDIA FreeBSD driver that is affected by this
1004 problem (i.e., the only driver to use either version 1.8 or 1.9 of the
1005 NV-CONTROL X extension) is 1.0-8756.
1008 ______________________________________________________________________________
1010 Chapter 8. Specifying OpenGL Environment Variable Settings
1011 ______________________________________________________________________________
1014 8A. FULL SCENE ANTIALIASING
1016 Antialiasing is a technique used to smooth the edges of objects in a scene to
1017 reduce the jagged "stairstep" effect that sometimes appears. By setting the
1018 appropriate environment variable, you can enable full-scene antialiasing in
1019 any OpenGL application on these GPUs.
1021 Several antialiasing methods are available and you can select between them by
1022 setting the __GL_FSAA_MODE environment variable appropriately. Note that
1023 increasing the number of samples taken during FSAA rendering may decrease
1026 To see the available values for __GL_FSAA_MODE along with their descriptions,
1029 nvidia-settings --query=fsaa --verbose
1031 The __GL_FSAA_MODE environment variable uses the same integer values that are
1032 used to configure FSAA through nvidia-settings and the NV-CONTROL X extension.
1035 8B. ANISOTROPIC TEXTURE FILTERING
1037 Automatic anisotropic texture filtering can be enabled by setting the
1038 environment variable __GL_LOG_MAX_ANISO. The possible values are:
1040 __GL_LOG_MAX_ANISO Filtering Type
1041 ---------------------------------- ----------------------------------
1042 0 No anisotropic filtering
1043 1 2x anisotropic filtering
1044 2 4x anisotropic filtering
1045 3 8x anisotropic filtering
1046 4 16x anisotropic filtering
1048 4x and greater are only available on GeForce3 or newer GPUs; 16x is only
1049 available on GeForce 6800 or newer GPUs.
1054 Setting the environment variable __GL_SYNC_TO_VBLANK to a non-zero value will
1055 force glXSwapBuffers to sync to your monitor's vertical refresh (perform a
1056 swap only during the vertical blanking period).
1058 When using __GL_SYNC_TO_VBLANK with TwinView, OpenGL can only sync to one of
1059 the display devices; this may cause tearing corruption on the display device
1060 to which OpenGL is not syncing. You can use the environment variable
1061 __GL_SYNC_DISPLAY_DEVICE to specify to which display device OpenGL should
1062 sync. You should set this environment variable to the name of a display
1063 device; for example "CRT-1". Look for the line "Connected display device(s):"
1064 in your X log file for a list of the display devices present and their names.
1065 You may also find it useful to review Chapter 10 "Configuring Twinview" and
1066 the section on Ensuring Identical Mode Timings in Chapter 16.
1069 8D. CONTROLLING THE SORTING OF OPENGL FBCONFIGS
1071 The NVIDIA GLX implementation sorts FBConfigs returned by glXChooseFBConfig()
1072 as described in the GLX specification. To disable this behavior set
1073 __GL_SORT_FBCONFIGS to 0 (zero), then FBConfigs will be returned in the order
1074 they were received from the X server. To examine the order in which FBConfigs
1075 are returned by the X server run:
1077 nvidia-settings --glxinfo
1079 This option may be be useful to work around problems in which applications
1080 pick an unexpected FBConfig.
1083 8E. OPENGL YIELD BEHAVIOR
1085 There are several cases where the NVIDIA OpenGL driver needs to wait for
1086 external state to change before continuing. To avoid consuming too much CPU
1087 time in these cases, the driver will sometimes yield so the kernel can
1088 schedule other processes to run while the driver waits. For example, when
1089 waiting for free space in a command buffer, if the free space has not become
1090 available after a certain number of iterations, the driver will yield before
1091 it continues to loop.
1093 By default, the driver calls sched_yield() to do this. However, this can cause
1094 the calling process to be scheduled out for a relatively long period of time
1095 if there are other, same-priority processes competing for time on the CPU. One
1096 example of this is when an OpenGL-based composite manager is moving and
1097 repainting a window and the X server is trying to update the window as it
1098 moves, which are both CPU-intensive operations.
1100 You can use the __GL_YIELD environment variable to work around these
1101 scheduling problems. This variable allows the user to specify what the driver
1102 should do when it wants to yield. The possible values are:
1105 --------------- ------------------------------------------------------
1106 <unset> By default, OpenGL will call sched_yield() to yield.
1107 "NOTHING" OpenGL will never yield.
1108 "USLEEP" OpenGL will call usleep(0) to yield.
1112 8F. CONTROLLING WHICH OPENGL FBCONFIGS ARE AVAILABLE
1114 The NVIDIA GLX implementation will hide FBConfigs that are associated with a
1115 32-bit ARGB visual when the XLIB_SKIP_ARGB_VISUALS environment variable is
1116 defined. This matches the behavior of libX11, which will hide those visuals
1117 from XGetVisualInfo and XMatchVisualInfo. This environment variable is useful
1118 when applications are confused by the presence of these FBConfigs.
1120 ______________________________________________________________________________
1122 Chapter 9. Configuring AGP
1123 ______________________________________________________________________________
1125 There are several choices for configuring the NVIDIA kernel module's use of
1126 AGP: you can choose to either use the NVIDIA AGP module (NVAGP), or the AGP
1127 module that comes with the FreeBSD kernel (AGPGART). This is controlled
1128 through the "NvAGP" option in your X config file:
1130 Option "NvAgp" "0" ... disables AGP support
1131 Option "NvAgp" "1" ... use NVAGP, if possible
1132 Option "NvAgp" "2" ... use AGPGART, if possible
1133 Option "NvAGP" "3" ... try AGPGART; if that fails, try NVAGP
1135 Unlike other operating systems such as Linux, this option is not the only
1136 controlling factor at this point; because of known problems, 'nvidia.ko' is
1137 built without support for FreeBSD's AGP driver by default. This behavior can
1138 be changed, see 'nv-freebsd.h' for details.
1140 Note that if you built nvidia.ko with support for the FreeBSD driver it will
1141 not load unless 'agp.ko' is loaded. 'agp.ko' is special in that you can not
1142 load it after the system boot is complete, you need to append the following
1143 line to '/boot/loader.conf' to make sure it is pre-loaded:
1145 # -- load FreeBSD AGP GART driver -- #
1148 Also note that if 'agp.ko' is loaded, it could conflict with the NVIDIA AGP
1149 GART driver (NvAGP), resulting in stability problems; for this reason, the
1150 NVIDIA driver will abort NvAGP initialization when it detects 'agp.ko'.
1152 Current FreeBSD releases are shipped with 'agp.ko' built into the kernel; in
1153 order to allow NvAGP to work, the kernel can be rebuilt without 'device agp'
1154 or the following entry added to '/boot/device.hints':
1156 hint.agp.0.disabled="1"
1158 When built with support for the FreeBSD AGP driver, 'nvidia.ko' will fall back
1159 to using NvAGP when it doesn't detect 'agp.ko' (this will be the case when
1160 'agp.ko' does not support your AGP chipset or was explicitly disabled with
1163 It is highly recommended that you use the NVIDIA AGP driver.
1165 The following AGP chipsets are supported by the NVIDIA AGP driver; for all
1166 other chipsets it is recommended that you use the AGPGART module.
1168 Supported AGP Chipsets
1169 ----------------------------------------------------------------------
1173 Intel 815 ("Solano")
1174 Intel 820 ("Camino")
1176 Intel 840 ("Carmel")
1177 Intel 845 ("Brookdale")
1179 Intel 850 ("Tehama")
1181 Intel 860 ("Colusa")
1182 Intel 865G ("Springdale")
1183 Intel 875P ("Canterwood")
1184 Intel E7205 ("Granite Bay")
1185 Intel E7505 ("Placer")
1186 AMD 751 ("Irongate")
1204 Micron SAMDDR ("Samurai")
1205 Micron SCIDDR ("Scimitar")
1234 If you are experiencing AGP stability problems, you should be aware of the
1237 Additional AGP Information
1239 AGP drive strength BIOS setting (Via-based motherboards)
1241 Many Via-based motherboards allow adjusting the AGP drive strength in the
1242 system BIOS. The setting of this option largely affects system stability,
1243 the range between 0xEA and 0xEE seems to work best for NVIDIA hardware.
1244 Setting either nibble to 0xF generally results in severe stability
1247 If you decide to experiment with this, you need to be aware of the fact
1248 that you are doing so at your own risk and that you may render your system
1249 unbootable with improper settings until you reset the setting to a working
1250 value (w/ a PCI graphics card or by resetting the BIOS to its default
1255 Make sure you have the latest system BIOS provided by the motherboard
1258 On ALi1541 and ALi1647 chipsets, NVIDIA drivers disable AGP to work around
1259 timing and signal integrity problems. You can force AGP to be enabled on
1260 these chipsets by setting NVreg_EnableALiAGP to 1. Note that this may
1261 cause the system to become unstable.
1263 Early system BIOS revisions for the ASUS A7V8X-X KT400 motherboard
1264 misconfigure the chipset when an AGP 2.x graphics card is installed; if X
1265 hangs on your ASUS KT400 system with NvAGP enabled and the installed
1266 graphics card is not an AGP 8x device, make sure that you have the latest
1267 system BIOS installed.
1270 ______________________________________________________________________________
1272 Chapter 10. Configuring TwinView
1273 ______________________________________________________________________________
1275 TwinView is a mode of operation where two display devices (digital flat
1276 panels, CRTs, and TVs) can display the contents of a single X screen in any
1277 arbitrary configuration. This method of multiple monitor use has several
1278 distinct advantages over other techniques (such as Xinerama):
1281 o A single X screen is used. The NVIDIA driver conceals all information
1282 about multiple display devices from the X server; as far as X is
1283 concerned, there is only one screen.
1285 o Both display devices share one frame buffer. Thus, all the functionality
1286 present on a single display (e.g., accelerated OpenGL) is available with
1289 o No additional overhead is needed to emulate having a single desktop.
1292 If you are interested in using each display device as a separate X screen, see
1296 10A. X CONFIG TWINVIEW OPTIONS
1298 To enable TwinView, you must specify the following option in the Device
1299 section of your X Config file:
1303 You may also use any of the following options, though they are not required:
1305 Option "MetaModes" "<list of MetaModes>"
1307 Option "SecondMonitorHorizSync" "<hsync range(s)>"
1308 Option "SecondMonitorVertRefresh" "<vrefresh range(s)>"
1310 Option "HorizSync" "<hsync range(s)>"
1311 Option "VertRefresh" "<vrefresh range(s)>"
1313 Option "TwinViewOrientation" "<relationship of head 1 to head 0>"
1314 Option "ConnectedMonitor" "<list of connected display devices>"
1316 See detailed descriptions of each option below.
1318 Alternatively, you can enable TwinView by running
1320 nvidia-xconfig --twinview
1322 and restarting your X server. Or, you can configure TwinView dynamically in
1323 the "Display Configuration" page in nvidia-settings.
1326 10B. DETAILED DESCRIPTION OF OPTIONS
1331 This option is required to enable TwinView; without it, all other TwinView
1332 related options are ignored.
1334 SecondMonitorHorizSync
1335 SecondMonitorVertRefresh
1337 You specify the constraints of the second monitor through these options.
1338 The values given should follow the same convention as the "HorizSync" and
1339 "VertRefresh" entries in the Monitor section. As the XF86Config man page
1340 explains it: the ranges may be a comma separated list of distinct values
1341 and/or ranges of values, where a range is given by two distinct values
1342 separated by a dash. The HorizSync is given in kHz, and the VertRefresh is
1345 These options are normally not needed: by default, the NVIDIA X driver
1346 retrieves the valid frequency ranges from the display device's EDID (see
1347 Appendix F for a description of the "UseEdidFreqs" option). The
1348 SecondMonitor options will override any frequency ranges retrieved from
1354 Which display device is "first" and which is "second" is often unclear.
1355 For this reason, you may use these options instead of the SecondMonitor
1356 versions. With these options, you can specify a semicolon-separated list
1357 of frequency ranges, each optionally prepended with a display device name.
1360 Option "HorizSync" "CRT-0: 50-110; DFP-0: 40-70"
1361 Option "VertRefresh" "CRT-0: 60-120; DFP-0: 60"
1363 See Appendix G on Display Device Names for more information.
1365 These options are normally not needed: by default, the NVIDIA X driver
1366 retrieves the valid frequency ranges from the display device's EDID (see
1367 Appendix F for a description of the "UseEdidFreqs" option). The
1368 "HorizSync" and "VertRefresh" options override any frequency ranges
1369 retrieved from the EDID or any frequency ranges specified with the
1370 "SecondMonitorHorizSync" and "SecondMonitorVertRefresh" options.
1374 MetaModes are "containers" that store information about what mode should
1375 be used on each display device at any given time. Even if only one display
1376 device is actively in use, the NVIDIA X driver always uses a MetaMode to
1377 encapsulate the mode information per display device, so that it can
1378 support dynamically enabling TwinView.
1380 Multiple MetaModes list the combinations of modes and the sequence in
1381 which they should be used. When the NVIDIA driver tells X what modes are
1382 available, it is really the minimal bounding box of the MetaMode that is
1383 communicated to X, while the "per display device" mode is kept internal to
1384 the NVIDIA driver. In MetaMode syntax, modes within a MetaMode are comma
1385 separated, and multiple MetaModes are separated by semicolons. For
1388 "<mode name 0>, <mode name 1>; <mode name 2>, <mode name 3>"
1390 Where <mode name 0> is the name of the mode to be used on display device 0
1391 concurrently with <mode name 1> used on display device 1. A mode switch
1392 will then cause <mode name 2> to be used on display device 0 and <mode
1393 name 3> to be used on display device 1. Here is an example MetaMode:
1395 Option "MetaModes" "1280x1024,1280x1024; 1024x768,1024x768"
1397 If you want a display device to not be active for a certain MetaMode, you
1398 can use the mode name "NULL", or simply omit the mode name entirely:
1400 "1600x1200, NULL; NULL, 1024x768"
1404 "1600x1200; , 1024x768"
1406 Optionally, mode names can be followed by offset information to control
1407 the positioning of the display devices within the virtual screen space;
1410 "1600x1200 +0+0, 1024x768 +1600+0; ..."
1412 Offset descriptions follow the conventions used in the X "-geometry"
1413 command line option; i.e., both positive and negative offsets are valid,
1414 though negative offsets are only allowed when a virtual screen size is
1415 explicitly given in the X config file.
1417 When no offsets are given for a MetaMode, the offsets will be computed
1418 following the value of the TwinViewOrientation option (see below). Note
1419 that if offsets are given for any one of the modes in a single MetaMode,
1420 then offsets will be expected for all modes within that single MetaMode;
1421 in such a case offsets will be assumed to be +0+0 when not given.
1423 When not explicitly given, the virtual screen size will be computed as the
1424 the bounding box of all MetaMode bounding boxes. MetaModes with a bounding
1425 box larger than an explicitly given virtual screen size will be discarded.
1427 A MetaMode string can be further modified with a "Panning Domain"
1428 specification; e.g.,
1430 "1024x768 @1600x1200, 800x600 @1600x1200"
1432 A panning domain is the area in which a display device's viewport will be
1433 panned to follow the mouse. Panning actually happens on two levels with
1434 TwinView: first, an individual display device's viewport will be panned
1435 within its panning domain, as long as the viewport is contained by the
1436 bounding box of the MetaMode. Once the mouse leaves the bounding box of
1437 the MetaMode, the entire MetaMode (i.e., all display devices) will be
1438 panned to follow the mouse within the virtual screen. Note that individual
1439 display devices' panning domains default to being clamped to the position
1440 of the display devices' viewports, thus the default behavior is just that
1441 viewports remain "locked" together and only perform the second type of
1444 The most beneficial use of panning domains is probably to eliminate dead
1445 areas -- regions of the virtual screen that are inaccessible due to
1446 display devices with different resolutions. For example:
1448 "1600x1200, 1024x768"
1450 produces an inaccessible region below the 1024x768 display. Specifying a
1451 panning domain for the second display device:
1453 "1600x1200, 1024x768 @1024x1200"
1455 provides access to that dead area by allowing you to pan the 1024x768
1456 viewport up and down in the 1024x1200 panning domain.
1458 Offsets can be used in conjunction with panning domains to position the
1459 panning domains in the virtual screen space (note that the offset
1460 describes the panning domain, and only affects the viewport in that the
1461 viewport must be contained within the panning domain). For example, the
1462 following describes two modes, each with a panning domain width of 1900
1463 pixels, and the second display is positioned below the first:
1465 "1600x1200 @1900x1200 +0+0, 1024x768 @1900x768 +0+1200"
1467 Because it is often unclear which mode within a MetaMode will be used on
1468 each display device, mode descriptions within a MetaMode can be prepended
1469 with a display device name. For example:
1471 "CRT-0: 1600x1200, DFP-0: 1024x768"
1473 If no MetaMode string is specified, then the X driver uses the modes
1474 listed in the relevant "Display" subsection, attempting to place matching
1475 modes on each display device.
1479 This option controls the positioning of the second display device relative
1480 to the first within the virtual X screen, when offsets are not explicitly
1481 given in the MetaModes. The possible values are:
1483 "RightOf" (the default)
1489 When "Clone" is specified, both display devices will be assigned an offset
1492 Because it is often unclear which display device is "first" and which is
1493 "second", TwinViewOrientation can be confusing. You can further clarify
1494 the TwinViewOrientation with display device names to indicate which
1495 display device is positioned relative to which display device. For
1498 "CRT-0 LeftOf DFP-0"
1503 With this option you can override what the NVIDIA kernel module detects is
1504 connected to your graphics card. This may be useful, for example, if any
1505 of your display devices do not support detection using Display Data
1506 Channel (DDC) protocols. Valid values are a comma-separated list of
1507 display device names; for example:
1513 WARNING: this option overrides what display devices are detected by the
1514 NVIDIA kernel module, and is very seldom needed. You really only need this
1515 if a display device is not detected, either because it does not provide
1516 DDC information, or because it is on the other side of a KVM
1517 (Keyboard-Video-Mouse) switch. In most other cases, it is best not to
1518 specify this option.
1521 Just as in all X config entries, spaces are ignored and all entries are case
1525 10C. DYNAMIC TWINVIEW
1527 Using the NV-CONTROL X extension, the display devices in use by an X screen,
1528 the mode pool for each display device, and the MetaModes for each X screen can
1529 be dynamically manipulated. The "Display Configuration" page in
1530 nvidia-settings uses this functionality to modify the MetaMode list and then
1531 uses XRandR to switch between MetaModes. This gives the ability to dynamically
1534 The details of how this works are documented in the nv-control-dpy.c sample
1535 NV-CONTROL client in the nvidia-settings source tarball.
1537 Because the NVIDIA X driver can now transition into and out of TwinView
1538 dynamically, MetaModes are always used internally by the NVIDIA X driver,
1539 regardless of how many display devices are currently in use by the X screen
1540 and regardless of whether the TwinView X configuration option was specified.
1542 One implication of this implementation is that each MetaMode must be uniquely
1543 identifiable to the XRandR X extension. Unfortunately, two MetaModes with the
1544 same bounding box will look the same to XRandR. For example, two MetaModes
1545 with different orientations:
1547 "CRT: 1600x1200 +0+0, DFP: 1600x1200 +1600+0"
1548 "CRT: 1600x1200 +1600+0, DFP: 1600x1200 +0+0"
1550 will look identical to the XRandR or XF86VidMode X extensions, because they
1551 have the same total size (3200x1200), and nvidia-settings would not be able to
1552 use XRandR to switch between these MetaModes. To work around this limitation,
1553 the NVIDIA X driver "lies" about the refresh rate of each MetaMode, using the
1554 refresh rate of the MetaMode as a unique identifier.
1556 The XRandR extension is currently being redesigned by the X.Org community, so
1557 the refresh rate workaround may be removed at some point in the future. This
1558 workaround can also be disabled by setting the "DynamicTwinView" X
1559 configuration option to FALSE, which will disable NV-CONTROL support for
1560 manipulating MetaModes, but will cause the XRandR and XF86VidMode visible
1561 refresh rate to be accurate.
1564 FREQUENTLY ASKED TWINVIEW QUESTIONS
1566 Q. Nothing gets displayed on my second monitor; what is wrong?
1568 A. Monitors that do not support monitor detection using Display Data Channel
1569 (DDC) protocols (this includes most older monitors) are not detectable by
1570 your NVIDIA card. You need to explicitly tell the NVIDIA X driver what you
1571 have connected using the "ConnectedMonitor" option; e.g.,
1573 Option "ConnectedMonitor" "CRT, CRT"
1577 Q. Will window managers be able to appropriately place windows (e.g., avoiding
1578 placing windows across both display devices, or in inaccessible regions of
1579 the virtual desktop)?
1581 A. Yes. The NVIDIA X driver provides a Xinerama extension that X clients (such
1582 as window managers) can use to discover the current TwinView configuration.
1583 Note that the Xinerama protocol provides no way to notify clients when a
1584 configuration change occurs, so if you modeswitch to a different MetaMode,
1585 your window manager will still think you have the previous configuration.
1586 Using the Xinerama extension, in conjunction with the XF86VidMode extension
1587 to get modeswitch events, window managers should be able to determine the
1588 TwinView configuration at any given time.
1590 Unfortunately, the data provided by XineramaQueryScreens() appears to
1591 confuse some window managers; to work around such broken window mangers,
1592 you can disable communication of the TwinView screen layout with the
1593 "NoTwinViewXineramaInfo" X config Option (see Appendix F for details).
1595 The order that display devices are reported in via the TwinView Xinerama
1596 information can be configured with the TwinViewXineramaInfoOrder X
1597 configuration option.
1599 Be aware that the NVIDIA driver cannot provide the Xinerama extension if
1600 the X server's own Xinerama extension is being used. Explicitly specifying
1601 Xinerama in the X config file or on the X server commandline will prohibit
1602 NVIDIA's Xinerama extension from installing, so make sure that the X
1603 server's log file does not contain:
1605 (++) Xinerama: enabled
1607 if you want the NVIDIA driver to be able to provide the Xinerama extension
1610 Another solution is to use panning domains to eliminate inaccessible
1611 regions of the virtual screen (see the MetaMode description above).
1613 A third solution is to use two separate X screens, rather than use
1614 TwinView. See Chapter 12.
1617 Q. Why can I not get a resolution of 1600x1200 on the second display device
1618 when using a GeForce2 MX?
1620 A. Because the second display device on the GeForce2 MX was designed to be a
1621 digital flat panel, the Pixel Clock for the second display device is only
1622 150 MHz. This effectively limits the resolution on the second display
1623 device to somewhere around 1280x1024 (for a description of how Pixel Clock
1624 frequencies limit the programmable modes, see the XFree86 Video Timings
1625 HOWTO). This constraint is not present on GeForce4 or GeForce FX GPUs --
1626 the maximum pixel clock is the same on both heads.
1629 Q. Do video overlays work across both display devices?
1631 A. Hardware video overlays only work on the first display device. The current
1632 solution is that blitted video is used instead on TwinView.
1635 Q. How are virtual screen dimensions determined in TwinView?
1637 A. After all requested modes have been validated, and the offsets for each
1638 MetaMode's viewports have been computed, the NVIDIA driver computes the
1639 bounding box of the panning domains for each MetaMode. The maximum bounding
1640 box width and height is then found.
1642 Note that one side effect of this is that the virtual width and virtual
1643 height may come from different MetaModes. Given the following MetaMode
1646 "1600x1200,NULL; 1024x768+0+0, 1024x768+0+768"
1648 the resulting virtual screen size will be 1600 x 1536.
1651 Q. Can I play full screen games across both display devices?
1653 A. Yes. While the details of configuration will vary from game to game, the
1654 basic idea is that a MetaMode presents X with a mode whose resolution is
1655 the bounding box of the viewports for that MetaMode. For example, the
1658 Option "MetaModes" "1024x768,1024x768; 800x600,800x600"
1659 Option "TwinViewOrientation" "RightOf"
1661 produce two modes: one whose resolution is 2048x768, and another whose
1662 resolution is 1600x600. Games such as Quake 3 Arena use the VidMode
1663 extension to discover the resolutions of the modes currently available. To
1664 configure Quake 3 Arena to use the above MetaMode string, add the following
1665 to your q3config.cfg file:
1667 seta r_customaspect "1"
1668 seta r_customheight "600"
1669 seta r_customwidth "1600"
1670 seta r_fullscreen "1"
1673 Note that, given the above configuration, there is no mode with a
1674 resolution of 800x600 (remember that the MetaMode "800x600, 800x600" has a
1675 resolution of 1600x600"), so if you change Quake 3 Arena to use a
1676 resolution of 800x600, it will display in the lower left corner of your
1677 screen, with the rest of the screen grayed out. To have single head modes
1678 available as well, an appropriate MetaMode string might be something like:
1680 "800x600,800x600; 1024x768,NULL; 800x600,NULL; 640x480,NULL"
1682 More precise configuration information for specific games is beyond the
1683 scope of this document, but the above examples coupled with numerous online
1684 sources should be enough to point you in the right direction.
1687 ______________________________________________________________________________
1689 Chapter 11. Configuring GLX in Xinerama
1690 ______________________________________________________________________________
1692 The NVIDIA FreeBSD Driver supports GLX when Xinerama is enabled on similar
1693 GPUs. The Xinerama extension takes multiple physical X screens (possibly
1694 spanning multiple GPUs), and binds them into one logical X screen. This allows
1695 windows to be dragged between GPUs and to span across multiple GPUs. The
1696 NVIDIA driver supports hardware accelerated OpenGL rendering across all NVIDIA
1697 GPUs when Xinerama is enabled.
1699 To configure Xinerama
1701 1. Configure multiple X screens (refer to the XF86Config(5x) or
1702 xorg.conf(5x) man pages for details).
1704 2. Enable Xinerama by adding the line
1706 Option "Xinerama" "True"
1708 to the "ServerFlags" section of your X config file.
1713 o Using identical GPUs is recommended. Some combinations of non-identical,
1714 but similar, GPUs are supported. If a GPU is incompatible with the rest
1715 of a Xinerama desktop then no OpenGL rendering will appear on the screens
1716 driven by that GPU. Rendering will still appear normally on screens
1717 connected to other supported GPUs. In this situation the X log file will
1718 include a message of the form:
1722 (WW) NVIDIA(2): The GPU driving screen 2 is incompatible with the rest of
1723 (WW) NVIDIA(2): the GPUs composing the desktop. OpenGL rendering will
1724 (WW) NVIDIA(2): be disabled on screen 2.
1728 o The NVIDIA X driver must be used for all X screens in the server.
1730 o Only the intersection of capabilities across all GPUs will be advertised.
1732 The maximum OpenGL viewport size depends on the hardware used, and is
1733 described by the following table. If an OpenGL window is larger than the
1734 maximum viewport, regions beyond the viewport will be blank.
1736 OpenGL Viewport Maximums in Xinerama
1738 GeForce GPUs before GeForce 8: 4096 x 4096 pixels
1739 GeForce 8 and newer GPUs: 8192 x 8192 pixels
1740 Quadro: as large as the Xinerama
1744 o X configuration options that affect GLX operation (e.g.: stereo,
1745 overlays) should be set consistently across all X screens in the X
1751 o Versions of XFree86 prior to 4.5 and versions of X.Org prior to 6.8.0
1752 lack the required interfaces to properly implement overlays with the
1753 Xinerama extension. On earlier server versions mixing overlays and
1754 Xinerama will result in rendering corruption. If you are using the
1755 Xinerama extension with overlays, it is recommended that you upgrade to
1756 XFree86 4.5, X.Org 6.8.0, or newer.
1759 ______________________________________________________________________________
1761 Chapter 12. Configuring Multiple X Screens on One Card
1762 ______________________________________________________________________________
1764 GPUs that support TwinView (Chapter 10) can also be configured to treat each
1765 connected display device as a separate X screen.
1767 While there are several disadvantages to this approach as compared to TwinView
1768 (e.g.: windows cannot be dragged between X screens, hardware accelerated
1769 OpenGL cannot span the two X screens), it does offer several advantages over
1772 o If each display device is a separate X screen, then properties that may
1773 vary between X screens may vary between displays (e.g.: depth, root
1776 o Hardware that can only be used on one display at a time (e.g.: video
1777 overlays, hardware accelerated RGB overlays), and which consequently
1778 cannot be used at all when in TwinView, can be exposed on the first X
1779 screen when each display is a separate X screen.
1781 o TwinView is a fairly new feature. X has historically used one screen per
1785 To configure two separate X screens to share one graphics card, here is what
1786 you will need to do:
1788 First, create two separate Device sections, each listing the BusID of the
1789 graphics card to be shared and listing the driver as "nvidia", and assign each
1793 Identifier "nvidia0"
1795 # Edit the BusID with the location of your graphics card
1801 Identifier "nvidia1"
1803 # Edit the BusID with the location of your graphics card
1808 Then, create two Screen sections, each using one of the Device sections:
1811 Identifier "Screen0"
1815 Subsection "Display"
1817 Modes "1600x1200" "1024x768" "800x600" "640x480"
1822 Identifier "Screen1"
1826 Subsection "Display"
1828 Modes "1600x1200" "1024x768" "800x600" "640x480"
1832 (Note: You'll also need to create a second Monitor section) Finally, update
1833 the ServerLayout section to use and position both Screen sections:
1835 Section "ServerLayout"
1838 Screen 1 "Screen1" leftOf "Screen0"
1842 For further details, refer to the XF86Config(5x) or xorg.conf(5x) man pages.
1844 ______________________________________________________________________________
1846 Chapter 13. Configuring TV-Out
1847 ______________________________________________________________________________
1849 NVIDIA GPU-based graphics cards with a TV-Out connector can use a television
1850 as another display device (the same way that it would use a CRT or digital
1851 flat panel). The TV can be used by itself, or in conjunction with another
1852 display device in a TwinView or multiple X screen configuration. If a TV is
1853 the only display device connected to your graphics card, it will be used as
1854 the primary display when you boot your system (i.e. the console will come up
1855 on the TV just as if it were a CRT).
1857 The NVIDIA X driver populates the mode pool for the TV with all the mode sizes
1858 that the driver supports with the given TV standard and the TV encoder on the
1859 graphics card. These modes are given names that correspond to their
1860 resolution; e.g., "800x600".
1862 Because these TV modes only depend on the TV encoder and the TV standard, TV
1863 modes do not go through normal mode validation. The X configuration options
1864 HorizSync and VertRefresh are not used for TV mode validation.
1866 Additionally, the NVIDIA driver contains a hardcoded list of mode sizes that
1867 it can drive for each combination of TV encoder and TV standard. Therefore,
1868 custom modelines in your X configuration file are ignored for TVs.
1870 To use your TV with X, there are several relevant X configuration options:
1872 o The Modes in the screen section of your X configuration file; you can use
1873 these to request any of the modes in the mode pool which the X driver
1874 created for this combination of TV standard and TV encoder. Examples
1875 include "640x480" and "800x600". If in doubt, use "nvidia-auto-select".
1877 o The "TVStandard" option should be added to your screen section; valid
1880 TVStandard Description
1881 ------------- --------------------------------------------------
1882 "PAL-B" used in Belgium, Denmark, Finland, Germany,
1883 Guinea, Hong Kong, India, Indonesia, Italy,
1884 Malaysia, The Netherlands, Norway, Portugal,
1885 Singapore, Spain, Sweden, and Switzerland
1886 "PAL-D" used in China and North Korea
1887 "PAL-G" used in Denmark, Finland, Germany, Italy,
1888 Malaysia, The Netherlands, Norway, Portugal,
1889 Spain, Sweden, and Switzerland
1890 "PAL-H" used in Belgium
1891 "PAL-I" used in Hong Kong and The United Kingdom
1892 "PAL-K1" used in Guinea
1893 "PAL-M" used in Brazil
1894 "PAL-N" used in France, Paraguay, and Uruguay
1895 "PAL-NC" used in Argentina
1896 "NTSC-J" used in Japan
1897 "NTSC-M" used in Canada, Chile, Colombia, Costa Rica,
1898 Ecuador, Haiti, Honduras, Mexico, Panama, Puerto
1899 Rico, South Korea, Taiwan, United States of
1900 America, and Venezuela
1901 "HD480i" 480 line interlaced
1902 "HD480p" 480 line progressive
1903 "HD720p" 720 line progressive
1904 "HD1080i" 1080 line interlaced
1905 "HD1080p" 1080 line progressive
1906 "HD576i" 576 line interlace
1907 "HD576p" 576 line progressive
1909 The line in your X config file should be something like:
1911 Option "TVStandard" "NTSC-M"
1913 If you do not specify a TVStandard, or you specify an invalid value, the
1914 default "NTSC-M" will be used. Note: if your country is not in the above
1915 list, select the country closest to your location.
1917 o The "UseDisplayDevice" option can be used if there are multiple display
1918 devices connected, and you want the connected TV to be used instead of
1919 the connected CRTs and/or DFPs. E.g.,
1921 Option "UseDisplayDevice" "TV"
1923 Using the "UseDisplayDevice" option, rather than the "ConnectedMonitor"
1924 option, is recommended.
1926 o The "TVOutFormat" option can be used to force the output format. Without
1927 this option, the driver autodetects the output format. Unfortunately, it
1928 does not always do this correctly. The output format can be forced with
1929 the "TVOutFormat" option; valid values are:
1931 TVOutFormat Description Supported TV
1933 ------------------- ------------------- -------------------
1934 "AUTOSELECT" The driver PAL, NTSC, HD
1938 "COMPOSITE" Force Composite PAL, NTSC
1940 "SVIDEO" Force S-Video PAL, NTSC
1942 "COMPONENT" Force Component HD
1945 "SCART" Force Scart output PAL, NTSC
1949 The line in your X config file should be something like:
1951 Option "TVOutFormat" "SVIDEO"
1954 o The "TVOverScan" option can be used to enable Overscan, when the TV
1955 encoder supports it. Valid values are decimal values in the range 1.0
1956 (which means overscan as much as possible: make the image as large as
1957 possible) and 0.0 (which means disable overscanning: make the image as
1958 small as possible). Overscanning is disabled (0.0) by default.
1960 The NVIDIA X driver may not restore the console correctly with XFree86
1961 versions older than 4.3 when the console is a TV. This is due to binary
1962 incompatibilities between XFree86 int10 modules. If you use a TV as your
1963 console it is recommended that you upgrade to XFree86 4.3 or later.
1965 ______________________________________________________________________________
1967 Chapter 14. Using the XRandR Extension
1968 ______________________________________________________________________________
1970 X.Org version X11R6.8.1 contains support for the rotation component of the
1971 XRandR extension, which allows screens to be rotated at 90 degree increments.
1973 The driver supports rotation with the extension when 'Option "RandRRotation"'
1974 is enabled in the X config file.
1976 Workstation RGB or CI overlay visuals will function at lower performance and
1977 the video overlay will not be available when RandRRotation is enabled.
1979 You can query the available rotations using the 'xrandr' command line
1980 interface to the RandR extension by running:
1984 You can set the rotation orientation of the screen by running any of:
1991 Rotation may also be set through the nvidia-settings configuration utility in
1992 the "Rotation Settings" panel.
1994 SLI and rotation are incompatible. Rotation will be disabled when SLI is
1997 TwinView and rotation can be used together, but rotation affects the entire
1998 desktop. This means that the same rotation setting will apply to both display
1999 devices in a TwinView pair. Note also that the "TwinViewOrientation" option
2000 applies before rotation does. For example, if you have two screens
2001 side-by-side and you want to rotate them, you should set "TwinViewOrientation"
2002 to "Above" or "Below".
2004 ______________________________________________________________________________
2006 Chapter 15. Configuring a Notebook
2007 ______________________________________________________________________________
2010 15A. INSTALLATION AND CONFIGURATION
2012 Installation and configuration of the NVIDIA FreeBSD Driver Set on a notebook
2013 is the same as for any desktop environment, with a few additions, as described
2017 15B. POWER MANAGEMENT
2019 All notebook NVIDIA GPUs support power management, both S3 (also known as
2020 "Standby" or "Suspend to RAM") and S4 (also known as "Hibernate", "Suspend to
2021 Disk" or "SWSUSP"). Power management is system-specific and is dependent upon
2022 all the components in the system; some systems may be more problematic than
2025 Most recent notebook NVIDIA GPUs also support PowerMizer, which monitors
2026 application work load to adjust system parameters to deliver the optimal
2027 balance of performance and battery life. However, PowerMizer is only enabled
2028 by default on some notebooks. Please see the known issues below for more
2032 15C. HOTKEY SWITCHING OF DISPLAY DEVICES
2034 Mobile NVIDIA GPUs also have the capacity to react to a display change hotkey
2035 event, toggling between each of the connected display devices and each
2036 possible combination of the connected display devices (note that only 2
2037 display devices may be active at a time).
2039 Hotkey switching dynamically changes the TwinView configuration; a given
2040 hotkey event will indicate which display devices should be in use at that
2041 time, and all MetaModes currently configured on the X screen will be updated
2042 to use the new configuration of display devices.
2044 Another important aspect of hotkey functionality is that you can dynamically
2045 connect and remove display devices to/from your notebook and use the hotkey to
2046 activate and deactivate them without restarting X.
2048 Note that there are two approaches to implementing this hotkey support: ACPI
2051 Most recent notebooks use ACPI events to deliver hotkeys from the System BIOS
2052 to the graphics driver. This is the preferred method of delivering hotkey
2053 events, but is still a new feature under most UNIX platforms and may not
2054 always function correctly.
2056 The polling mechanism requires checking during the vertical blanking interval
2057 for a hotkey status change. It is an older mechanism for handling hotkeys, and
2058 is therefore not supported on all notebooks and is not tested by notebook
2059 manufacturers. It also does not always report the same combinations of display
2060 devices that are reported by ACPI hotkey events.
2062 The NVIDIA FreeBSD Driver will attempt to use ACPI hotkey events, if possible.
2063 In the case that ACPI hotkey event support is not available, the driver will
2064 revert back to trying hotkey polling. In the case that the notebook does not
2065 support hotkey polling, hotkeys will not work. Please see the known issues
2066 section below for more details.
2068 When switching away from X to a virtual terminal, the VGA console will always
2069 be restored to the display device on which it was present when X was started.
2070 Similarly, when switching back into X, the same display device configuration
2071 will be used as when you switched away, regardless of what display change
2072 hotkey activity occurred while the virtual terminal was active.
2077 All notebook NVIDIA GPUs support docking, however support may be limited by
2078 the OS or system. There are three types of notebook docking (hot, warm, and
2079 cold), which refer to the state of the system when the docking event occurs.
2080 hot refers to a powered on system with a live desktop, warm refers to a system
2081 that has entered a suspended power management state, and cold refers to a
2082 system that has been powered off. Only warm and cold docking are supported by
2088 All notebook NVIDIA GPUs support TwinView. TwinView on a notebook can be
2089 configured in the same way as on a desktop computer (refer to Chapter 10 );
2090 note that in a TwinView configuration using the notebook's internal flat panel
2091 and an external CRT, the CRT is the primary display device (specify its
2092 HorizSync and VertRefresh in the Monitor section of your X config file) and
2093 the flat panel is the secondary display device (specify its HorizSync and
2094 VertRefresh through the SecondMonitorHorizSync and SecondMonitorVertRefresh
2097 The "UseEdidFreqs" X config option is enabled by default, so normally you
2098 should not need to specify the "SecondMonitorHorizSync" and
2099 "SecondMonitorVertRefresh" options. See the description of the UseEdidFreqs
2100 option in Appendix F for details).
2103 15F. KNOWN NOTEBOOK ISSUES
2105 There are a few known issues associated with notebooks:
2107 o Display change hotkey switching is not available on all notebooks. In
2108 some cases, the ACPI infrastructure is not fully supported by the NVIDIA
2109 FreeBSD Driver. Work is ongoing to increase the robustness of NVIDIA's
2110 support in this area. Toshiba and Lenovo notebooks are known to be
2113 o ACPI Display change hotkey switching is not supported by X.Org X servers
2114 earlier than 1.2.0; see EnableACPIHotkeys in Appendix F for details.
2116 o In many cases, suspending and/or resuming will fail. As mentioned above,
2117 this functionality is very system-specific. There are still many cases
2118 that are problematic. Here are some tips that may help:
2120 o In some cases, hibernation can have bad interactions with the PCI
2121 Express bus clocks, which can lead to system hangs when entering
2122 hibernation. This issue is still being investigated, but a known
2123 workaround is to leave an OpenGL application running when
2127 o On some notebooks, PowerMizer is not enabled by default. This issue is
2128 being investigated, and there is no known workaround.
2130 o ACPI is not currently supported on FreeBSD As a result, ACPI hotkey
2131 events are not supported.
2133 o The video overlay only works on the first display device on which you
2134 started X. For example, if you start X on the internal LCD, run a video
2135 application that uses the video overlay (uses the "Video Overlay" adapter
2136 advertised through the XV extension), and then hotkey switch to add a
2137 second display device, the video will not appear on the second display
2138 device. To work around this, you can either configure the video
2139 application to use the "Video Blitter" adapter advertised through the XV
2140 extension (this is always available), or hotkey switch to the display
2141 device on which you want to use the video overlay *before* starting X.
2144 ______________________________________________________________________________
2146 Chapter 16. Programming Modes
2147 ______________________________________________________________________________
2149 The NVIDIA Accelerated FreeBSD Graphics Driver supports all standard VGA and
2150 VESA modes, as well as most user-written custom mode lines; double-scan modes
2151 are supported on all hardware. Interlaced modes are supported on all GeForce
2152 FX/Quadro FX and newer GPUs, and certain older GPUs; the X log file will
2153 contain a message "Interlaced video modes are supported on this GPU" if
2154 interlaced modes are supported.
2156 To request one or more standard modes for use in X, you can simply add a
2157 "Modes" line such as:
2159 Modes "1600x1200" "1024x768" "640x480"
2161 in the appropriate Display subsection of your X config file (see the
2162 XF86Config(5x) or xorg.conf(5x) man pages for details). Or, the
2163 nvidia-xconfig(1) utility can be used to request additional modes; for
2166 nvidia-xconfig --mode 1600x1200
2168 See the nvidia-xconfig(1) man page for details.
2171 16A. DEPTH, BITS PER PIXEL, AND PITCH
2173 While not directly a concern when programming modes, the bits used per pixel
2174 is an issue when considering the maximum programmable resolution; for this
2175 reason, it is worthwhile to address the confusion surrounding the terms
2176 "depth" and "bits per pixel". Depth is how many bits of data are stored per
2177 pixel. Supported depths are 8, 15, 16, and 24. Most video hardware, however,
2178 stores pixel data in sizes of 8, 16, or 32 bits; this is the amount of memory
2179 allocated per pixel. When you specify your depth, X selects the bits per pixel
2180 (bpp) size in which to store the data. Below is a table of what bpp is used
2181 for each possible depth:
2184 ---------------------------------- ----------------------------------
2190 Lastly, the "pitch" is how many bytes in the linear frame buffer there are
2191 between one pixel's data, and the data of the pixel immediately below. You can
2192 think of this as the horizontal resolution multiplied by the bytes per pixel
2193 (bits per pixel divided by 8). In practice, the pitch may be more than this
2194 product due to alignment constraints.
2197 16B. MAXIMUM RESOLUTIONS
2199 The NVIDIA Accelerated FreeBSD Graphics Driver and NVIDIA GPU-based graphics
2200 cards support resolutions up to 8192x8192 pixels for the GeForce 8 series and
2201 above, and up to 4096x4096 pixels for the GeForce 7 series and below, though
2202 the maximum resolution your system can support is also limited by the amount
2203 of video memory (see USEFUL FORMULAS for details) and the maximum supported
2204 resolution of your display device (monitor/flat panel/television). Also note
2205 that while use of a video overlay does not limit the maximum resolution or
2206 refresh rate, video memory bandwidth used by a programmed mode does affect the
2210 16C. USEFUL FORMULAS
2212 The maximum resolution is a function both of the amount of video memory and
2213 the bits per pixel you elect to use:
2215 HR * VR * (bpp/8) = Video Memory Used
2217 In other words, the amount of video memory used is equal to the horizontal
2218 resolution (HR) multiplied by the vertical resolution (VR) multiplied by the
2219 bytes per pixel (bits per pixel divided by eight). Technically, the video
2220 memory used is actually the pitch times the vertical resolution, and the pitch
2221 may be slightly greater than (HR * (bpp/8)) to accommodate the hardware
2222 requirement that the pitch be a multiple of some value.
2224 Note that this is just memory usage for the frame buffer; video memory is also
2225 used by other things, such as OpenGL and pixmap caching.
2227 Another important relationship is that between the resolution, the pixel clock
2228 (also known as the dot clock) and the vertical refresh rate:
2230 RR = PCLK / (HFL * VFL)
2232 In other words, the refresh rate (RR) is equal to the pixel clock (PCLK)
2233 divided by the total number of pixels: the horizontal frame length (HFL)
2234 multiplied by the vertical frame length (VFL) (note that these are the frame
2235 lengths, and not just the visible resolutions). As described in the XFree86
2236 Video Timings HOWTO, the above formula can be rewritten as:
2238 PCLK = RR * HFL * VFL
2240 Given a maximum pixel clock, you can adjust the RR, HFL and VFL as desired, as
2241 long as the product of the three is consistent. The pixel clock is reported in
2242 the log file. Your X log should contain a line like this:
2244 (--) NVIDIA(0): ViewSonic VPD150 (DFP-1): 165 MHz maximum pixel clock
2246 which indicates the maximum pixel clock for that display device.
2249 16D. HOW MODES ARE VALIDATED
2251 In traditional XFree86/X.Org mode validation, the X server takes as a starting
2252 point the X server's internal list of VESA standard modes, plus any modes
2253 specified with special ModeLines in the X configuration file's Monitor
2254 section. These modes are validated against criteria such as the valid
2255 HorizSync/VertRefresh frequency ranges for the user's monitor (as specified in
2256 the Monitor section of the X configuration file), as well as the maximum pixel
2259 Once the X server has determined the set of valid modes, it takes the list of
2260 user requested modes (i.e., the set of modes named in the "Modes" line in the
2261 Display subsection of the Screen section of X configuration file), and finds
2262 the "best" validated mode with the requested name.
2264 The NVIDIA X driver uses a variation on the above approach to perform mode
2265 validation. During X server initialization, the NVIDIA X driver builds a pool
2266 of valid modes for each display device. It gathers all possible modes from
2269 o The display device's EDID
2271 o The X server's built-in list
2273 o Any user-specified ModeLines in the X configuration file
2275 o The VESA standard modes
2277 For every possible mode, the mode is run through mode validation. The core of
2278 mode validation is still performed similarly to traditional XFree86/X.Org mode
2279 validation: the mode timings are checked against things such as the valid
2280 HorizSync and VertRefresh ranges and the maximum pixelclock. Note that each
2281 individual stage of mode validation can be independently controlled through
2282 the "ModeValidation" X configuration option.
2284 Note that when validating interlaced mode timings, VertRefresh specifies the
2285 field rate, rather than the frame rate. For example, the following modeline
2286 has a vertical refresh rate of 87 Hz:
2289 # 1024x768i @ 87Hz (industry standard)
2290 ModeLine "1024x768" 44.9 1024 1032 1208 1264 768 768 776 817 +hsync +vsync
2294 Invalid modes are discarded; valid modes are inserted into the mode pool. See
2295 MODE VALIDATION REPORTING for how to get more details on mode validation
2296 results for each considered mode.
2298 Valid modes are given a unique name that is guaranteed to be unique across the
2299 whole mode pool for this display device. This mode name is constructed
2300 approximately like this:
2302 <width>x<height>_<refreshrate>
2304 (e.g., "1600x1200_85")
2306 The name may also be prepended with another number to ensure the mode is
2307 unique; e.g., "1600x1200_85_0".
2309 As validated modes are inserted into the mode pool, duplicate modes are
2310 removed, and the mode pool is sorted, such that the "best" modes are at the
2311 beginning of the mode pool. The sorting is based roughly on:
2315 o Source (EDID-provided modes are prioritized higher than VESA-provided
2316 modes, which are prioritized higher than modes that were in the X
2317 server's built-in list)
2321 Once modes from all mode sources are validated and the mode pool is
2322 constructed, all modes with the same resolution are compared; the best mode
2323 with that resolution is added to the mode pool a second time, using just the
2324 resolution as its unique modename (e.g., "1600x1200"). In this way, when you
2325 request a mode using the traditional names (e.g., "1600x1200"), you still get
2326 what you got before (the 'best' 1600x1200 mode); the added benefit is that all
2327 modes in the mode pool can be addressed by a unique name.
2329 When verbose logging is enabled (see the FAQ section on increasing the amount
2330 of data printed in the X log file), the mode pool for each display device is
2331 printed to the X log file.
2333 After the mode pool is built for all display devices, the requested modes (as
2334 specified in the X configuration file), are looked up from the mode pool. Each
2335 requested mode that can be matched against a mode in the mode pool is then
2336 advertised to the X server and is available to the user through the X server's
2337 mode switching hotkeys (ctrl-alt-plus/minus) and the XRandR and XF86VidMode X
2340 If only one display device is in use by the X screen when the X server starts,
2341 all modes in the mode pool are implicitly made available to the X server. See
2342 the "IncludeImplicitMetaModes" X configuration option in Appendix F for
2346 16E. THE NVIDIA-AUTO-SELECT MODE
2348 You can request a special mode by name in the X config file, named
2349 "nvidia-auto-select". When the X driver builds the mode pool for a display
2350 device, it selects one of the modes as the "nvidia-auto-select" mode; a new
2351 entry is made in the mode pool, and "nvidia-auto-select" is used as the unique
2354 The "nvidia-auto-select" mode is intended to be a reasonable mode for the
2355 display device in question. For example, the "nvidia-auto-select" mode is
2356 normally the native resolution for flatpanels, as reported by the flatpanel's
2357 EDID, or one of the detailed timings from the EDID. The "nvidia-auto-select"
2358 mode is guaranteed to always be present, and to always be defined as something
2359 considered valid by the X driver for this display device.
2361 Note that the "nvidia-auto-select" mode is not necessarily the largest
2362 possible resolution, nor is it necessarily the mode with the highest refresh
2363 rate. Rather, the "nvidia-auto-select" mode is selected such that it is a
2364 reasonable default. The selection process is roughly:
2367 o If the EDID for the display device reported a preferred mode timing, and
2368 that mode timing is considered a valid mode, then that mode is used as
2369 the "nvidia-auto-select" mode. You can check if the EDID reported a
2370 preferred timing by starting X with logverbosity greater than or equal to
2371 5 (see the FAQ section on increasing the amount of data printed in the X
2372 log file), and looking at the EDID printout; if the EDID contains a line:
2374 Prefer first detailed timing : Yes
2376 Then the first mode listed under the "Detailed Timings" in the EDID will
2379 o If the EDID did not provide a preferred timing, the best detailed timing
2380 from the EDID is used as the "nvidia-auto-select" mode.
2382 o If the EDID did not provide any detailed timings (or there was no EDID at
2383 all), the best valid mode not larger than 1024x768 is used as the
2384 "nvidia-auto-select" mode. The 1024x768 limit is imposed here to restrict
2385 use of modes that may have been validated, but may be too large to be
2386 considered a reasonable default, such as 2048x1536.
2388 o If all else fails, the X driver will use a built-in 800 x 600 60Hz mode
2389 as the "nvidia-auto-select" mode.
2392 If no modes are requested in the X configuration file, or none of the
2393 requested modes can be found in the mode pool, then the X driver falls back to
2394 the "nvidia-auto-select" mode, so that X can always start. Appropriate warning
2395 messages will be printed to the X log file in these fallback scenarios.
2397 You can add the "nvidia-auto-select" mode to your X configuration file by
2400 nvidia-xconfig --mode nvidia-auto-select
2402 and restarting your X server.
2404 The X driver can generally do a much better job of selecting the
2405 "nvidia-auto-select" mode if the display device's EDID is available. This is
2406 one reason why the "IgnoreEDID" X configuration option has been deprecated,
2407 and that it is recommended to only use the "UseEDID" X configuration option
2408 sparingly. Note that, rather than globally disable all uses of the EDID with
2409 the "UseEDID" option, you can individually disable each particular use of the
2410 EDID using the "UseEDIDFreqs", "UseEDIDDpi", and/or the "NoEDIDModes" argument
2411 in the "ModeValidation" X configuration option.
2414 16F. MODE VALIDATION REPORTING
2416 When log verbosity is set to 6 or higher (see FAQ
2417 section on increasing the amount of data printed in the X log file), the X log
2418 will record every mode that is considered for each display device's mode pool,
2419 and report whether the mode passed or failed. For modes that were considered
2420 invalid, the log will report why the mode was considered invalid.
2423 16G. ENSURING IDENTICAL MODE TIMINGS
2425 Some functionality, such as Active Stereo with TwinView, requires control over
2426 exactly which mode timings are used. For explicit control over which mode
2427 timings are used on each display device, you can specify the ModeLine you want
2428 to use (using one of the ModeLine generators available), and using a unique
2429 name. For example, if you wanted to use 1024x768 at 120 Hz on each monitor in
2430 TwinView with active stereo, you might add something like this to the monitor
2431 section of your X configuration file:
2433 # 1024x768 @ 120.00 Hz (GTF) hsync: 98.76 kHz; pclk: 139.05 MHz
2434 Modeline "1024x768_120" 139.05 1024 1104 1216 1408 768 769 772 823
2437 Then, in the Screen section of your X config file, specify a MetaMode like
2440 Option "MetaModes" "1024x768_120, 1024x768_120"
2444 16H. ADDITIONAL INFORMATION
2446 An XFree86 ModeLine generator, conforming to the GTF Standard is available at
2447 http://gtf.sourceforge.net/. Additional generators can be found by searching
2448 for "modeline" on freshmeat.net.
2450 ______________________________________________________________________________
2452 Chapter 17. Configuring Flipping and UBB
2453 ______________________________________________________________________________
2455 The NVIDIA Accelerated FreeBSD Graphics Driver supports Unified Back Buffer
2456 (UBB) and OpenGL Flipping. These features can provide performance gains in
2459 o Unified Back Buffer (UBB): UBB is available only on the Quadro family of
2460 GPUs (Quadro4 NVS excluded) and is enabled by default when there is
2461 sufficient video memory available. This can be disabled with the UBB X
2462 config option described in Appendix F. When UBB is enabled, all windows
2463 share the same back, stencil and depth buffers. When there are many
2464 windows, the back, stencil and depth usage will never exceed the size of
2465 that used by a full screen window. However, even for a single small
2466 window, the back, stencil, and depth video memory usage is that of a full
2467 screen window. In that case video memory may be used less efficiently
2468 than in the non-UBB case.
2470 o Flipping: When OpenGL flipping is enabled, OpenGL can perform buffer
2471 swaps by changing which buffer the DAC scans out rather than copying the
2472 back buffer contents to the front buffer; this is generally a much higher
2473 performance mechanism and allows tearless swapping during the vertical
2474 retrace (when __GL_SYNC_TO_VBLANK is set). The conditions under which
2475 OpenGL can flip are slightly complicated, but in general: on GeForce or
2476 newer hardware, OpenGL can flip when a single full screen unobscured
2477 OpenGL application is running, and __GL_SYNC_TO_VBLANK is enabled.
2478 Additionally, OpenGL can flip on Quadro hardware even when an OpenGL
2479 window is partially obscured or not full screen or __GL_SYNC_TO_VBLANK is
2483 ______________________________________________________________________________
2485 Appendix C. The Sysctl Interface
2486 ______________________________________________________________________________
2488 The sysctl interface allows you to obtain run-time information about the
2489 driver, any installed NVIDIA graphics cards and the AGP status. It also allows
2490 you to control low-level configuration options and/or overrides.
2492 The various pieces of information are held in a hierarchy under hw.nvidia and
2493 are accessible with the sysctl(8) command.
2495 NVIDIA sysctl Entries
2499 Prints the installed driver revision
2503 These OIDs provide information about NVIDIA device 'n':
2506 -------------------------------- --------------------------------
2507 model the device's product name
2508 irq the IRQ claimed by this device
2509 vbios the device's Video BIOS revision
2510 type the bus type of this device
2513 hw.nvidia.agp.host-bridge.*
2514 hw.nvidia.agp.card.*
2516 These OIDs provide information about the AGP capabilities of the installed
2517 AGP graphics card and host-bridge respectively. These values are most
2518 likely to be correct after system boot and before the X server is started
2519 (and the AGP subsystem initialized).
2522 -------------- ---------------------------------------------------
2523 rates the AGP rates supported by this device
2524 fw if the device supports AGP fast-writes
2525 sba if the device supports AGP side-band-addressing
2526 registers the device's AGP registers, status:command
2529 hw.nvidia.agp.status.*
2531 Prints AGP status information based on the AGP command registers of the
2532 host-bridge and of the AGP card.
2535 -------------- ---------------------------------------------------
2536 status if AGP is enabled or disabled
2537 driver which driver is being used
2538 rate the programmed AGP rate
2539 fw if fast-writes are enabled or disabled
2540 sba if side-band-addressing is enabled or disabled
2543 hw.nvidia.registry.*
2545 Low-level kernel module configuration options. Changing these is typically
2546 not necessary and potentially dangerous. If you do need to change any of
2547 these options, you will need to do so BEFORE you start the X server.
2550 -------------- ---------------------------------------------------
2551 status if AGP is enabled or disabled
2552 driver which driver is being used
2553 rate the programmed AGP rate
2554 fw if fast-writes are enabled or disabled
2555 sba if side-band-addressing is enabled or disabled
2559 ______________________________________________________________________________
2561 Appendix D. Configuring Low-level Parameters
2562 ______________________________________________________________________________
2564 The NVIDIA resource manager recognizes several low-level configuration
2565 parameters that can be set using the sysctl driver interface BEFORE the X
2566 server is started. Normally you should not need to modify any of these
2567 parameters, but it is sometimes necessary or desirable to do so.
2569 To view the current settings of these parameters, you need to issue this
2570 sysctl command ('nvidia.ko' needs to be loaded):
2572 % sysctl -a hw.nvidia.registry
2574 To change any of the parameters, you need to pass the complete name of the OID
2575 followed by '=' and the new value, e.g.:
2577 % sysctl hw.nvidia.registry.EnableVia4x=1
2579 It is possible to automate setting these parameters by adding them to the
2580 '/etc/sysctl.conf' file. See `man 5 sysctl.conf` for details.
2582 The following parameters are recognized by 'nvidia.ko':
2584 Resource Manager Parameters
2586 VideoMemoryTypeOverride
2588 We normally detect memory type on TNT cards by scanning the embedded BIOS.
2589 Unfortunately, we've seen some cases where a TNT card has been flashed
2590 with the wrong BIOS. For example, an SDRAM based TNT has been flashed with
2591 an SGRAM BIOS, and therefore claims to be an SGRAM TNT. We've therefore
2592 provided an override here. Make sure to set the value toe the type of
2593 memory used on your card.
2596 -------------------------------- --------------------------------
2600 Note that we can only do so much here. There are border cases where even
2601 this fails. For example, if 2 TNT cards are in the same system, one SGRAM,
2604 This option is disabled by default, see below for information on how to
2609 We've had problems with some Via chipsets in 4x mode, we need force them
2610 back down to 2x mode. If you'd like to experiment with retaining 4x mode,
2611 you may try setting this value to 1 If that hangs the system, you're stuck
2612 with 2x mode; there's nothing we can do about it.
2615 -------------- ---------------------------------------------------
2616 0 disable AGP 4x on Via chipsets (default)
2617 1 enable AGP 4x on Via chipsets
2622 Some ALi chipsets (ALi1541, ALi1647) are known to cause severe system
2623 stability problems with AGP enabled. To avoid lockups, we disable AGP on
2624 systems with these chipsets by default. It appears that updating the
2625 system BIOS and using recent versions of the kernel AGP Gart driver can
2626 make such systems much more stable. If you own a system with one of the
2627 aforementioned chipsets and had it working reasonably well previously, or
2628 if you want to experiment with BIOS and AGPGART revisions, you can
2629 re-enable AGP support by setting this option to 1.
2632 -------------- ---------------------------------------------------
2633 0 disable AGP on Ali1541 and ALi1647 (default)
2634 1 enable AGP on Ali1541 and ALi1647
2639 This options controls which AGP GART driver is used when no explicit
2640 request is made to change the default (X server).
2643 -------------- ---------------------------------------------------
2644 0 disable AGP support
2645 1 use the NVIDIA builtin driver (if possible)
2646 2 use the kernel's AGPGART driver (if possible)
2647 3 use any available driver (try 2, then 1)
2649 Note that the NVIDIA internal AGP GART driver will not be used if AGPGART
2650 was either statically linked into your kernel or built as a kernel module
2651 and loaded before the NVIDIA kernel module.
2655 Normally, the driver will compare speed modes of the chipset and the card,
2656 picking the highest common rate. This key forces a maximum limit, to limit
2657 the driver to lower speeds. The driver will not attempt a speed beyond
2658 what the chipset and card claim they are capable of.
2660 Make sure you really know what you're doing before you enable this
2661 override. By default, AGP drivers will enable the fastest AGP rate your
2662 card and motherboard chipset are capable of. Then, in some cases, our
2663 driver will force this rate down to work around bugs in both our chipsets,
2664 and motherboard chipsets. Using this variable will override our bug fixes.
2665 This may be desirable in some cases, but not most. THIS IS COMPLETELY
2668 This option expects a bitmask (7 = 1|2|3|4, 3=1|2, etc.)
2670 This option is disabled by default, see below for information on how to
2675 For stability reasons, the driver will not use Side Band Addressing even
2676 if both the host chipset and the AGP card support it. You may override
2677 this behavior with the following registry key. THIS IS COMPLETELY
2681 -------------- ---------------------------------------------------
2682 0 disable Side Band Addressing (default on x86, see
2684 1 enable Side Band Addressing (if supported)
2689 Similar to Side Band Addressing, Fast Writes are disabled by default. If
2690 you wish to enable them on systems that support them, you can do so with
2691 this registry key. Note that this may render your system unstable with
2692 many AGP chipsets. THIS IS COMPLETELY UNSUPPORTED!
2695 -------------------------------- --------------------------------
2696 0 disable Fast Writes (default)
2697 1 enable Fast Writes
2701 ______________________________________________________________________________
2703 Chapter 18. Using the X Composite Extension
2704 ______________________________________________________________________________
2706 X.Org X servers, beginning with X11R6.8.0, contain experimental support for a
2707 new X protocol extension called Composite. This extension allows windows to be
2708 drawn into pixmaps instead of directly onto the screen. In conjunction with
2709 the Damage and Render extensions, this allows a program called a composite
2710 manager to blend windows together to draw the screen.
2712 Performance will be degraded significantly if the "RenderAccel" option is
2713 disabled in xorg.conf. See Appendix F for more details.
2715 When the NVIDIA X driver is used with an X.Org X server X11R6.9.0 or newer and
2716 the Composite extension is enabled, NVIDIA's OpenGL implementation interacts
2717 properly with the Damage and Composite X extensions. This means that OpenGL
2718 rendering is drawn into offscreen pixmaps and the X server is notified of the
2719 Damage event when OpenGL renders to the pixmap. This allows OpenGL
2720 applications to behave properly in a composited X desktop.
2722 If the Composite extension is enabled on an X server older than X11R6.9.0,
2723 then GLX will be disabled. You can force GLX on while Composite is enabled on
2724 pre-X11R6.9.0 X servers with the "AllowGLXWithComposite" X configuration
2725 option. However, GLX will not render correctly in this environment. Upgrading
2726 your X server to X11R6.9.0 or newer is recommended.
2728 You can enable the Composite X extension by running 'nvidia-xconfig
2729 --composite'. Composite can be disabled with 'nvidia-xconfig --no-composite'.
2730 See the nvidia-xconfig(1) man page for details.
2732 If you are using Composite with GLX, it is recommended that you also enable
2733 the "DamageEvents" X option for enhanced performance. If you are using an
2734 OpenGL-based composite manager, you may also need the "DisableGLXRootClipping"
2735 option to obtain proper output.
2737 The Composite extension also causes problems with other driver components:
2739 o In X servers prior to X.Org 7.1, Xv cannot draw into pixmaps that have
2740 been redirected offscreen and will draw directly onto the screen instead.
2741 For some programs you can work around this issue by using an alternative
2742 video driver. For example, "mplayer -vo x11" will work correctly, as will
2743 "xine -V xshm". If you must use Xv with an older server, you can also
2744 disable the compositing manager and re-enable it when you are finished.
2746 On X.Org 7.1 and higher, the driver will properly redirect video into
2747 offscreen pixmaps. Note that the Xv adaptors will ignore the
2748 sync-to-vblank option when drawing into a redirected window.
2750 o Workstation overlays, stereo visuals, and the unified back buffer (UBB)
2751 are incompatible with Composite. These features will be automatically
2752 disabled when Composite is detected.
2755 This NVIDIA FreeBSD supports OpenGL rendering to 32-bit ARGB windows on X.Org
2756 7.2 and higher or when the "AddARGBGLXVisuals" X config file option is
2757 enabled. If you are an application developer, you can use these new visuals in
2758 conjunction with a composite manager to create translucent OpenGL
2762 GLX_RENDER_TYPE, GLX_RGBA_BIT,
2763 GLX_DRAWABLE_TYPE, GLX_WINDOW_BIT,
2768 GLX_DOUBLEBUFFER, True,
2771 GLXFBConfig *fbconfigs, fbconfig;
2772 int numfbconfigs, render_event_base, render_error_base;
2773 XVisualInfo *visinfo;
2774 XRenderPictFormat *pictFormat;
2776 /* Make sure we have the RENDER extension */
2777 if(!XRenderQueryExtension(dpy, &render_event_base, &render_error_base)) {
2778 fprintf(stderr, "No RENDER extension found\n");
2782 /* Get the list of FBConfigs that match our criteria */
2783 fbconfigs = glXChooseFBConfig(dpy, scrnum, attrib, &numfbconfigs);
2789 /* Find an FBConfig with a visual that has a RENDER picture format that
2791 for (i = 0; i < numfbconfigs; i++) {
2792 visinfo = glXGetVisualFromFBConfig(dpy, fbconfigs[i]);
2793 if (!visinfo) continue;
2794 pictFormat = XRenderFindVisualFormat(dpy, visinfo->visual);
2795 if (!pictFormat) continue;
2797 if(pictFormat->direct.alphaMask > 0) {
2798 fbconfig = fbconfigs[i];
2805 if (i == numfbconfigs) {
2806 /* None of the FBConfigs have alpha. Use a normal (opaque)
2807 * FBConfig instead */
2808 fbconfig = fbconfigs[0];
2809 visinfo = glXGetVisualFromFBConfig(dpy, fbconfig);
2810 pictFormat = XRenderFindVisualFormat(dpy, visinfo->visual);
2816 When rendering to a 32-bit window, keep in mind that the X RENDER extension,
2817 used by most composite managers, expects "premultiplied alpha" colors. This
2818 means that if your color has components (r,g,b) and alpha value a, then you
2819 must render (a*r, a*g, a*b, a) into the target window.
2821 More information about Composite can be found at
2822 http://freedesktop.org/Software/CompositeExt
2824 ______________________________________________________________________________
2826 Chapter 19. Using the nvidia-settings Utility
2827 ______________________________________________________________________________
2829 A graphical configuration utility, 'nvidia-settings', is included with the
2830 NVIDIA FreeBSD graphics driver. After installing the driver and starting X,
2831 you can run this configuration utility by running:
2835 in a terminal window.
2837 Detailed information about the configuration options available are documented
2838 in the help window in the utility.
2840 For more information, see the nvidia-settings man page or the user guide
2842 ftp://download.nvidia.com/XFree86/Linux-x86/nvidia-settings-user-guide.txt
2844 The source code to nvidia-settings is released as GPL and is available here:
2845 ftp://download.nvidia.com/XFree86/nvidia-settings/
2847 If you have trouble running the nvidia-settings binary shipped with the NVIDIA
2848 FreeBSD Graphics Driver, refer to the nvidia-settings entry in Chapter 6.
2850 ______________________________________________________________________________
2852 Chapter 20. Configuring SLI and Multi-GPU FrameRendering
2853 ______________________________________________________________________________
2855 The NVIDIA FreeBSD driver contains support for NVIDIA SLI FrameRendering and
2856 NVIDIA Multi-GPU FrameRendering. Both of these technologies allow an OpenGL
2857 application to take advantage of multiple GPUs to improve visual performance.
2859 The distinction between SLI and Multi-GPU is straightforward. SLI is used to
2860 leverage the processing power of GPUs across two or more graphics cards, while
2861 Multi-GPU is used to leverage the processing power of two GPUs colocated on
2862 the same graphics card. If you want to link together separate graphics cards,
2863 you should use the "SLI" X config option. Likewise, if you want to link
2864 together GPUs on the same graphics card, you should use the "MultiGPU" X
2865 config option. If you have two cards, each with two GPUs, and you wish to link
2866 them all together, you should use the "SLI" option.
2868 In FreeBSD, with two GPUs SLI and Multi-GPU can both operate in one of three
2869 modes: Alternate Frame Rendering (AFR), Split Frame Rendering (SFR), and
2870 Antialiasing (AA). When AFR mode is active, one GPU draws the next frame while
2871 the other one works on the frame after that. In SFR mode, each frame is split
2872 horizontally into two pieces, with one GPU rendering each piece. The split
2873 line is adjusted to balance the load between the two GPUs. AA mode splits
2874 antialiasing work between the two GPUs. Both GPUs work on the same scene and
2875 the result is blended together to produce the final frame. This mode is useful
2876 for applications that spend most of their time processing with the CPU and
2877 cannot benefit from AFR.
2879 With four GPUs, the same options are applicable. AFR mode cycles through all
2880 four GPUs, each GPU rendering a frame in turn. SFR mode splits the frame
2881 horizontally into four pieces. AA mode splits the work between the four GPUs,
2882 allowing antialiasing up to 64x. With four GPUs SLI can also operate in an
2883 additional mode, Alternate Frame Rendering of Antialiasing. (AFR of AA). With
2884 AFR of AA, pairs of GPUs render alternate frames, each GPU in a pair doing
2885 half of the antialiasing work. Note that these scenarios apply whether you
2886 have four separate cards or you have two cards, each with two GPUs.
2888 Multi-GPU is enabled by setting the "MultiGPU" option in the X configuration
2889 file; see Appendix F for details about the "MultiGPU" option.
2891 The nvidia-xconfig utility can be used to set the "MultiGPU" option, rather
2892 than modifying the X configuration file by hand. For example:
2894 % nvidia-xconfig --multigpu=on
2897 SLI is enabled by setting the "SLI" option in the X configuration file; see
2898 Appendix F for details about the SLI option.
2900 The nvidia-xconfig utility can be used to set the SLI option, rather than
2901 modifying the X configuration file by hand. For example:
2903 % nvidia-xconfig --sli=on
2907 20A. HARDWARE REQUIREMENTS
2909 SLI functionality requires:
2911 o Identical PCI-Express graphics cards
2913 o A supported motherboard
2915 o In most cases, a video bridge connecting the two graphics cards
2917 For the latest in supported SLI and Multi-GPU configurations, including SLI-
2918 and Multi-GPU capable GPUs and SLI-capable motherboards, see
2919 http://www.slizone.com.
2922 20B. OTHER NOTES AND REQUIREMENTS
2924 The following other requirements apply to SLI and Multi-GPU:
2926 o Mobile GPUs are NOT supported
2928 o SLI on Quadro-based graphics cards always requires a video bridge
2930 o TwinView is also not supported with SLI or Multi-GPU. Only one display
2931 can be used when SLI or Multi-GPU is enabled.
2933 o If X is configured to use multiple screens and screen 0 has SLI or
2934 Multi-GPU enabled, the other screens will be disabled. Note that if SLI
2935 or Multi-GPU is enabled, the GPUs used by that configuration will be
2936 unavailable for single GPU rendering.
2940 FREQUENTLY ASKED SLI AND MULTI-GPU QUESTIONS
2942 Q. Why is glxgears slower when SLI or Multi-GPU is enabled?
2944 A. When SLI or Multi-GPU is enabled, the NVIDIA driver must coordinate the
2945 operations of all GPUs when each new frame is swapped (made visible). For
2946 most applications, this GPU synchronization overhead is negligible.
2947 However, because glxgears renders so many frames per second, the GPU
2948 synchronization overhead consumes a significant portion of the total time,
2949 and the framerate is reduced.
2952 Q. Why is Doom 3 slower when SLI or Multi-GPU is enabled?
2954 A. The NVIDIA Accelerated FreeBSD Graphics Driver does not automatically
2955 detect the optimal SLI or Multi-GPU settings for games such as Doom 3 and
2956 Quake 4. To work around this issue, the environment variable __GL_DOOM3 can
2957 be set to tell OpenGL that Doom 3's optimal settings should be used. In
2958 Bash, this can be done in the same command that launches Doom 3 so the
2959 environment variable does not remain set for other OpenGL applications
2960 started in the same session:
2962 % __GL_DOOM3=1 doom3
2964 Doom 3's startup script can also be modified to set this environment
2968 # Needed to make symlinks/shortcuts work.
2969 # the binaries must run with correct working directory
2970 cd "/usr/local/games/doom3/"
2971 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:.
2973 exec ./doom.x86 "$@"
2975 This environment variable is temporary and will be removed in the future.
2978 ______________________________________________________________________________
2980 Chapter 21. Configuring Frame Lock and Genlock
2981 ______________________________________________________________________________
2983 NOTE: Frame Lock and Genlock features are supported only on specific hardware,
2986 Visual computing applications that involve multiple displays, or even multiple
2987 windows within a display, can require special signal processing and
2988 application controls in order to function properly. For example, in order to
2989 produce quality video recording of animated graphics, the graphics display
2990 must be synchronized with the video camera. As another example, applications
2991 presented on multiple displays must be synchronized in order to complete the
2992 illusion of a larger, virtual canvas.
2994 This synchronization is enabled through the frame lock and genlock
2995 capabilities of the NVIDIA driver. This section describes the setup and use of
2996 frame lock and genlock.
2999 21A. DEFINITION OF TERMS
3001 GENLOCK: Genlock refers to the process of synchronizing the pixel scanning of
3002 one or more displays to an external synchronization source. NVIDIA Genlock
3003 requires the external signal to be either TTL or composite, such as used for
3004 NTSC, PAL, or HDTV. It should be noted that the NVIDIA Genlock implementation
3005 is guaranteed only to be frame-synchronized, and not necessarily
3008 FRAME LOCK: Frame Lock involves the use of hardware to synchronize the frames
3009 on each display in a connected system. When graphics and video are displayed
3010 across multiple monitors, frame locked systems help maintain image continuity
3011 to create a virtual canvas. Frame lock is especially critical for stereo
3012 viewing, where the left and right fields must be in sync across all displays.
3014 In short, to enable genlock means to sync to an external signal. To enable
3015 frame lock means to sync 2 or more display devices to a signal generated
3016 internally by the hardware, and to use both means to sync 2 or more display
3017 devices to an external signal.
3019 SWAP SYNC: Swap sync refers to the synchronization of buffer swaps of multiple
3020 application windows. By means of swap sync, applications running on multiple
3021 systems can synchronize the application buffer swaps between all the systems.
3022 In order to work across multiple systems, swap sync requires that the systems
3025 G-SYNC DEVICE: A G-Sync Device refers to devices capable of Frame
3026 lock/Genlock. This can be a graphics card (Quadro FX 3000G) or a stand alone
3027 device (Quadro FX G-Sync). See "Supported Hardware" below.
3030 21B. SUPPORTED HARDWARE
3032 Frame lock and genlock are supported for the following hardware:
3035 ----------------------------------------------------------------------
3037 Quadro FX G-Sync, used in conjunction with a Quadro FX 4400, Quadro FX
3038 4500, or Quadro FX 5500
3039 Quadro FX G-Sync II, used in conjunction with a Quadro FX 4600, or Quadro
3046 Before you begin, you should check that your hardware has been properly
3047 installed. If you are using the Quadro FX 3000G, the genlock/frame lock signal
3048 processing hardware is located on the dual-slot card itself, and after
3049 installing the card, no additional setup is necessary.
3051 If you are using the Quadro FX G-Sync card in conjunction with a graphics
3052 card, the following additional setup steps are required. These steps must be
3053 performed when the system is off.
3055 1. On the Quadro FX G-Sync card, locate the fourteen-pin connector labeled
3056 "primary". If the associated ribbon cable is not already joined to this
3057 connector, do so now. If you plan to use frame lock or genlock in
3058 conjunction with SLI FrameRendering or Multi-GPU FrameRendering (see
3059 Chapter 20) or other multi-GPU configurations, you should connect the
3060 fourteen-pin connector labeled "secondary" to the second GPU. A section
3061 at the end of this appendix describes restrictions on such setups.
3063 2. Install the Quadro FX G-Sync card in any available slot. Note that the
3064 slot itself is only used for support, so even a known "bad" slot is
3065 acceptable. The slot must be close enough to the graphics card that the
3066 ribbon cable can reach.
3068 3. Connect the other end of the ribbon cable to the fourteen-pin connector
3069 on the graphics card.
3071 You may now boot the system and begin the software setup of genlock and/or
3072 frame lock. These instructions assume that you have already successfully
3073 installed the NVIDIA Accelerated FreeBSD Driver Set. If you have not done so,
3077 21D. CONFIGURATION WITH NVIDIA-SETTINGS GUI
3079 Frame lock and genlock are configured through the nvidia-settings utility. See
3080 the 'nvidia-settings(1)' man page, and the nvidia-settings online help (click
3081 the "Help" button in the lower right corner of the interface for per-page help
3084 From the nvidia-settings frame lock panel, you may control the addition of
3085 G-Sync (and display) devices to the frame lock/genlock group, monitor the
3086 status of that group, and enable/disable frame lock and genlock.
3088 After the system has booted and X Windows has been started, run
3093 You may wish to start this utility before continuing, as we refer to it
3094 frequently in the subsequent discussion.
3096 The setup of genlock and frame lock are described separately. We then describe
3097 the use of genlock and frame lock together.
3102 After the system has been booted, connect the external signal to the house
3103 sync connector (the BNC connector) on either the graphics card or the G-Sync
3104 card. There is a status LED next to the connector. A solid red LED indicates
3105 that the hardware cannot detect the timing signal. A green LED indicates that
3106 the hardware is detecting a timing signal. An occasional red flash is okay.
3107 The G-Sync device (graphics card or G-Sync card) will need to be configured
3108 correctly for the signal to be detected.
3110 In the frame lock panel of the nvidia-settings interface, add the X Server
3111 that contains the display and G-Sync devices that you would like to sync to
3112 this external source by clicking the "Add Devices..." button. An X Server is
3113 typically specified in the format "system:m", e.g.:
3115 mycomputer.domain.com:0
3121 After adding an X Server, rows will appear in the "G-Sync Devices" section on
3122 the frame lock panel that displays relevant status information about the
3123 G-Sync devices, GPUs attached to those G-Sync devices and the display devices
3124 driven by those GPUs. In particular, the G-Sync rows will display the server
3125 name and G-Sync device number along with "Receiving" LED, "Rate", "House" LED,
3126 "Port 0"/"Port 1" Images, and "Delay" information. The GPU rows will display
3127 the GPU product name information along with the GPU ID for the server. The
3128 Display Device rows will show the display device name and device type along
3129 with server/client check boxes, refresh rate, "Timing" LED and "Stereo" LED.
3131 Once the G-Sync and display devices have been added to the frame lock/genlock
3132 group, a Server display device will need to be selected. This is done by
3133 selecting the "Server" check box of the desired display device.
3135 If you are using a G-Sync card, you must also click the "Use House Sync if
3136 Present" check box. To enable synchronization of this G-Sync device to the
3137 external source, click the "Enable Frame Lock" button. The display device(s)
3138 may take a moment to stabilize. If it does not stabilize, you may have
3139 selected a synchronization signal that the system cannot support. You should
3140 disable synchronization by clicking the "Disable Frame Lock" button and check
3141 the external sync signal.
3143 Modifications to genlock settings (e.g., "Use House Sync if Present", "Add
3144 Devices...") must be done while synchronization is disabled.
3147 21F. FRAME LOCK SETUP
3149 Frame Lock is supported across an arbitrary number of Quadro FX 3000 or Quadro
3150 FX G-Sync systems, although mixing the two in the same frame lock group is not
3151 supported. Additionally, each system to be included in the frame lock group
3152 must be configured with identical mode timings. See Chapter 16 for information
3155 Connect the systems through their RJ45 ports using standard CAT5 patch cables.
3156 These ports are located on the frame lock card itself (either the Quadro FX
3157 3000 or the Quadro FX G-Sync card). DO NOT CONNECT A FRAME LOCK PORT TO AN
3158 ETHERNET CARD OR HUB. DOING SO MAY PERMANENTLY DAMAGE THE HARDWARE. The
3159 connections should be made in a daisy-chain fashion: each card has two RJ45
3160 ports, call them 1 and 2. Connect port 1 of system A to port 2 of system B,
3161 connect port 1 of system B to port 2 of system C, etc. Note that you will
3162 always have two empty ports in your frame lock group.
3164 The ports self-configure as inputs or outputs once frame lock is enabled. Each
3165 port has a yellow and a green LED that reflect this state. A flashing yellow
3166 LED indicates an output and a flashing green LED indicates an input. A solid
3167 green LED indicates that the port has not yet configured.
3169 In the frame lock panel of the nvidia-settings interface, add the X server
3170 that contains the display devices that you would like to include in the frame
3171 lock group by clicking the "Add Devices..." button (see the description for
3172 adding display devices in the previous section on GENLOCK SETUP. Like the
3173 genlock status indicators, the "Port 0" and "Port 1" columns in the table on
3174 the frame lock panel contain indicators whose states mirror the states of the
3175 physical LEDs on the RJ45 ports. Thus, you may monitor the status of these
3176 ports from the software interface.
3178 Any X Server can be added to the frame lock group, provided that
3180 1. The system supporting the X Server is configured to support frame lock
3181 and is connected via RJ45 cable to the other systems in the frame lock
3184 2. The system driving nvidia-settings can locate and has display privileges
3185 on the X server that is to be included for frame lock.
3187 A system can gain display privileges on a remote system by executing
3191 on the remote system. See the xhost(1) man page for details. Typically, frame
3192 lock is controlled through one of the systems that will be included in the
3193 frame lock group. While this is not a requirement, note that nvidia-settings
3194 will only display the frame lock panel when running on an X server that
3195 supports frame lock.
3197 To enable synchronization on these display devices, click the "Enable Frame
3198 Lock" button. The screens may take a moment to stabilize. If they do not
3199 stabilize, you may have selected mode timings that one or more of the systems
3200 cannot support. In this case you should disable synchronization by clicking
3201 the "Disable Frame Lock" button and refer to Chapter 16 for information on
3204 Modifications to frame lock settings (e.g. "Add/Remove Devices...") must be
3205 done while synchronization is disabled.
3207 nvidia-settings will not automatically enable Frame Lock via the
3208 nvidia-settings.rc file. To enable Frame Lock when starting the X server, a
3209 line such as the following can be added to the '~/.xinitrc' file:
3211 # nvidia-settings -a [gpu:0]/FrameLockEnable=1
3215 21G. FRAME LOCK + GENLOCK
3217 The use of frame lock and genlock together is a simple extension of the above
3218 instructions for using them separately. You should first follow the
3219 instructions for Frame Lock Setup, and then to one of the systems that will be
3220 included in the frame lock group, attach an external sync source. In order to
3221 sync the frame lock group to this single external source, you must select a
3222 display device driven by the GPU connected to the G-Sync card (through the
3223 primary connector) that is connected to the external source to be the signal
3224 server for the group. This is done by selecting the check box labeled "Server"
3225 of the tree on the frame lock panel in nvidia-settings. If you are using a
3226 G-Sync based frame lock group, you must also select the "Use House Sync if
3227 Present" check box. Enable synchronization by clicking the "Enable Frame Lock"
3228 button. As with other frame lock/genlock controls, you must select the signal
3229 server while synchronization is disabled.
3232 21H. CONFIGURATION WITH NVIDIA-SETTINGS COMMAND LINE
3234 Frame Lock may also be configured through the nvidia-settings command line.
3235 This method of configuring Frame Lock may be useful in a scripted environment
3236 to automate the setup process. (Note that the examples listed below depend on
3237 the actual hardware configuration and as such may not work as-is.)
3239 To properly configure Frame Lock, the following steps should be completed:
3241 1. Make sure Frame Lock Sync is disabled on all GPUs.
3243 2. Make sure all display devices that are to be frame locked have the same
3246 3. Configure which (display/GPU) device should be the master.
3248 4. Configure house sync (if applicable).
3250 5. Configure the slave display devices.
3252 6. Enable frame lock sync on the master GPU.
3254 7. Enable frame lock sync on the slave GPUs.
3256 8. Toggle the test signal on the master GPU (for testing the hardware
3260 For a full list of the nvidia-settings Frame Lock attributes, please see the
3261 'nvidia-settings(1)' man page. Examples:
3263 1. 1 System, 1 Frame Lock board, 1 GPU, and 1 display device syncing to the
3266 # - Make sure frame lock sync is disabled
3267 nvidia-settings -a [gpu:0]/FrameLockEnable=0
3268 nvidia-settings -q [gpu:0]/FrameLockEnable
3270 # - Query the enabled displays on the gpu
3271 nvidia-settings -q [gpu:0]/EnabledDisplays
3273 # - Check that the refresh rate is the one we want
3274 nvidia-settings -q [gpu:0]/RefreshRate
3276 # - Set the master display device to CRT-0. The desired display
3277 # device(s) to be set are passed in as a hexadecimal number
3278 # in which specific bits denote which display devices to set.
3281 # 0x00000001 - CRT-0
3282 # 0x00000002 - CRT-1
3283 # 0x00000003 - CRT-0 and CRT-1
3288 # 0x00020000 - DFP-1
3290 # 0x00010101 - CRT-0, TV-0 and DFP-0
3292 # 0x000000FF - All CRTs
3293 # 0x0000FF00 - All TVs
3294 # 0x00FF0000 - All DFPs
3296 # Note that the following command:
3298 # nvidia-settings -q [gpu:0]/EnabledDisplays
3300 # will list the available displays on the given GPU.
3302 nvidia-settings -a [gpu:0]/FrameLockMaster=0x00000001
3303 nvidia-settings -q [gpu:0]/FrameLockMaster
3305 # - Enable use of house sync signal
3306 nvidia-settings -a [framelock:0]/FrameLockUseHouseSync=1
3308 # - Configure the house sync signal video mode
3309 nvidia-settings -a [framelock:0]/FrameLockVideoMode=0
3311 # - Set the slave display device to none (to avoid
3312 # having unwanted display devices locked to the
3314 nvidia-settings -a [gpu:0]/FrameLockSlaves=0x00000000
3315 nvidia-settings -q [gpu:0]/FrameLockSlaves
3317 # - Enable frame lock
3318 nvidia-settings -a [gpu:0]/FrameLockEnable=1
3320 # - Toggle the test signal
3321 nvidia-settings -a [gpu:0]/FrameLockTestSignal=1
3322 nvidia-settings -a [gpu:0]/FrameLockTestSignal=0
3325 2. 2 Systems, each with 2 GPUs, 1 Frame Lock board and 1 display device per
3326 GPU syncing from the first system's first display device:
3328 # - Make sure frame lock sync is disabled
3329 nvidia-settings -a myserver:0[gpu:0]/FrameLockEnable=0
3330 nvidia-settings -a myserver:0[gpu:1]/FrameLockEnable=0
3331 nvidia-settings -a myslave1:0[gpu:0]/FrameLockEnable=0
3332 nvidia-settings -a myslave1:0[gpu:1]/FrameLockEnable=0
3334 # - Query the enabled displays on the GPUs
3335 nvidia-settings -q myserver:0[gpu:0]/EnabledDisplays
3336 nvidia-settings -q myserver:0[gpu:1]/EnabledDisplays
3337 nvidia-settings -q myslave1:0[gpu:0]/EnabledDisplays
3338 nvidia-settings -q myslave1:0[gpu:1]/EnabledDisplays
3340 # - Check the refresh rate is the same for all displays
3341 nvidia-settings -q myserver:0[gpu:0]/RefreshRate
3342 nvidia-settings -q myserver:0[gpu:1]/RefreshRate
3343 nvidia-settings -q myslave1:0[gpu:0]/RefreshRate
3344 nvidia-settings -q myslave1:0[gpu:1]/RefreshRate
3346 # - Make sure the display device we want as master is masterable
3347 nvidia-settings -q myserver:0[gpu:0]/FrameLockMasterable
3349 # - Set the master display device (CRT-0)
3350 nvidia-settings -a myserver:0[gpu:0]/FrameLockMaster=0x00000001
3352 # - Disable the house sync signal on the master device
3353 nvidia-settings -a myserver:0[framelock:0]/FrameLockUseHouseSync=0
3355 # - Set the slave display devices
3356 nvidia-settings -a myserver:0[gpu:1]/FrameLockSlaves=0x00000001
3357 nvidia-settings -a myslave1:0[gpu:0]/FrameLockSlaves=0x00000001
3358 nvidia-settings -a myslave1:0[gpu:1]/FrameLockSlaves=0x00000001
3360 # - Enable frame lock on server
3361 nvidia-settings -a myserver:0[gpu:0]/FrameLockEnable=1
3363 # - Enable frame lock on slave devices
3364 nvidia-settings -a myserver:0[gpu:1]/FrameLockEnable=1
3365 nvidia-settings -a myslave1:0[gpu:0]/FrameLockEnable=1
3366 nvidia-settings -a myslave1:0[gpu:1]/FrameLockEnable=1
3368 # - Toggle the test signal
3369 nvidia-settings -a myserver:0[gpu:0]/FrameLockTestSignal=1
3370 nvidia-settings -a myserver:0[gpu:0]/FrameLockTestSignal=0
3373 3. 1 System, 4 GPUs, 2 Frame Lock boards and 2 display devices per GPU
3374 syncing from the first GPU's display device:
3376 # - Make sure frame lock sync is disabled
3377 nvidia-settings -a [gpu:0]/FrameLockEnable=0
3378 nvidia-settings -a [gpu:1]/FrameLockEnable=0
3379 nvidia-settings -a [gpu:2]/FrameLockEnable=0
3380 nvidia-settings -a [gpu:3]/FrameLockEnable=0
3382 # - Query the enabled displays on the GPUs
3383 nvidia-settings -q [gpu:0]/EnabledDisplays
3384 nvidia-settings -q [gpu:1]/EnabledDisplays
3385 nvidia-settings -q [gpu:2]/EnabledDisplays
3386 nvidia-settings -q [gpu:3]/EnabledDisplays
3388 # - Check the refresh rate is the same for all displays
3389 nvidia-settings -q [gpu:0]/RefreshRate
3390 nvidia-settings -q [gpu:1]/RefreshRate
3391 nvidia-settings -q [gpu:2]/RefreshRate
3392 nvidia-settings -q [gpu:3]/RefreshRate
3394 # - Make sure the display device we want as master is masterable
3395 nvidia-settings -q myserver:0[gpu:0]/FrameLockMasterable
3397 # - Set the master display device (CRT-0)
3398 nvidia-settings -a [gpu:0]/FrameLockMaster=0x00000001
3400 # - Disable the house sync signal on the master device
3401 nvidia-settings -a [framelock:0]/FrameLockUseHouseSync=0
3403 # - Set the slave display devices
3404 nvidia-settings -a [gpu:0]/FrameLockSlaves=0x00000002 # CRT-1
3405 nvidia-settings -a [gpu:1]/FrameLockSlaves=0x00000003 # CRT-0 and CRT-1
3406 nvidia-settings -a [gpu:2]/FrameLockSlaves=0x00000003 # CRT-0 and CRT-1
3407 nvidia-settings -a [gpu:3]/FrameLockSlaves=0x00000003 # CRT-0 and CRT-1
3409 # - Enable frame lock on master GPU
3410 nvidia-settings -a [gpu:0]/FrameLockEnable=1
3412 # - Enable frame lock on slave devices
3413 nvidia-settings -a [gpu:1]/FrameLockEnable=1
3414 nvidia-settings -a [gpu:2]/FrameLockEnable=1
3415 nvidia-settings -a [gpu:3]/FrameLockEnable=1
3417 # - Toggle the test signal
3418 nvidia-settings -a [gpu:0]/FrameLockTestSignal=1
3419 nvidia-settings -a [gpu:0]/FrameLockTestSignal=0
3424 21I. LEVERAGING FRAME LOCK/GENLOCK IN OPENGL
3426 With the GLX_NV_swap_group extension, OpenGL applications can be implemented
3427 to join a group of applications within a system for local swap sync, and bind
3428 the group to a barrier for swap sync across a frame lock group. A universal
3429 frame counter is also provided to promote synchronization across applications.
3432 21J. FRAME LOCK RESTRICTIONS:
3434 The following restrictions must be met for enabling frame lock:
3436 1. All display devices set as client in a frame lock group must have the
3437 same mode timings as the server (master) display device. If a House Sync
3438 signal is used (instead of internal timings), all client display devices
3439 must be set to have the same refresh rate as the incoming house sync
3442 2. All X Screens (driving the selected client/server display devices) must
3443 have the same stereo setting. See Appendix F for instructions on how to
3444 set the stereo X option.
3446 3. The frame lock server (master) display device must be on a GPU on the
3447 primary connector to a G-Sync device.
3449 4. If connecting a single GPU to a G-Sync device, the primary connector must
3452 5. In configurations with more than one display device per GPU, we recommend
3453 enabling frame lock on all display devices on those GPUs.
3455 6. Virtual terminal switching or mode switching will disable frame lock on
3456 the display device. Note that the glXQueryFrameCountNV entry point
3457 (provided by the GLX_NV_swap_group extension) will only provide
3458 incrementing numbers while frame lock is enabled. Therefore, applications
3459 that use glXQueryFrameCountNV to control animation will appear to stop
3460 animating while frame lock is disabled.
3464 21K. SUPPORTED FRAME LOCK CONFIGURATIONS:
3466 The following configurations are currently supported:
3468 1. Basic Frame Lock: Single GPU, Single X Screen, Single Display Device with
3469 or without OpenGL applications that make use of Quad-Buffered Stereo
3470 and/or the GLX_NV_swap_group extension.
3472 2. Frame Lock + TwinView: Single GPU, Single X Screen, Multiple Display
3473 Devices with or without OpenGL applications that make use of
3474 Quad-Buffered Stereo and/or the GLX_NV_swap_group extension.
3476 3. Frame Lock + Xinerama: 1 or more GPU(s), Multiple X Screens, Multiple
3477 Display Devices with or without OpenGL applications that make use of
3478 Quad-Buffered Stereo and/or the GLX_NV_swap_group extension.
3480 4. Frame Lock + TwinView + Xinerama: 1 or more GPU(s), Multiple X Screens,
3481 Multiple Display Devices with or without OpenGL applications that make
3482 use of Quad-Buffered Stereo and/or the GLX_NV_swap_group extension.
3484 5. Frame Lock + SLI SFR, AFR, or AA: 2 GPUs, Single X Screen, Single Display
3485 Device with either OpenGL applications that make use of Quad-Buffered
3486 Stereo or the GLX_NV_swap_group extension. Note that for Frame Lock + SLI
3487 Frame Rendering applications that make use of both Quad-Buffered Stereo
3488 and the GLX_NV_swap_group extension are not supported. Note that only
3489 2-GPU SLI configurations are currently supported.
3491 6. Frame Lock + Multi-GPU SFR, AFR, or AA: 2 GPUs, Single X Screen, Single
3492 Display Device with either OpenGL applications that make use of
3493 Quad-Buffered Stereo or the GLX_NV_swap_group extension. Note that for
3494 Frame Lock + Multi-GPU Frame Rendering applications that make use of both
3495 Quad-Buffered Stereo and the GLX_NV_swap_group extension are not
3499 ______________________________________________________________________________
3501 Chapter 22. Configuring SDI Video Output
3502 ______________________________________________________________________________
3504 Broadcast, film, and video post production and digital cinema applications can
3505 require Serial Digital (SDI) or High Definition Serial Digital (HD-SDI) video
3506 output. SDI/HD-SDI is a digital video interface used for the transmission of
3507 uncompressed video signals as well as packetized data. SDI is standardized in
3508 ITU-R BT.656 and SMPTE 259M while HD-SDI is standardized in SMPTE 292M. SMPTE
3509 372M extends HD-SDI to define a dual-link configuration that uses a pair of
3510 SMPTE 292M links to provide a 2.970 Gbit/second interface. SMPTE 424M extends
3511 the interface further to define a single 2.97 Gbit/second serial data link.
3513 SDI and HD-SDI video output is provided through the use of the NVIDIA driver
3514 along with an NVIDIA SDI output daughter board. In addition to single- and
3515 dual-link SDI/HD-SDI digital video output, frame lock and genlock
3516 synchronization are provided in order to synchronize the outgoing video with
3517 an external source signal (see Chapter 21 for details on these technologies).
3518 This section describes the setup and use of the SDI video output.
3523 Before you begin, you should check that your hardware has been properly
3524 installed. If you are using the Quadro FX 4000 SDI, the SDI/HD-SDI hardware is
3525 located on the dual-slot card itself, and after installing the card, no
3526 additional setup is necessary. If you are using the Quadro FX 4500/5500 SDI or
3527 Quadro FX 4600/5600 SDI II, the following additional setup steps are required
3528 in order to connect the SDI daughter card to the graphics card. These steps
3529 must be performed when the system is off.
3531 1. Insert the NVIDIA SDI Output card into any available expansion slot
3532 within six inches of the NVIDIA Quadro graphics card. Secure the card's
3533 bracket using the method provided by the chassis manufacturer (usually a
3534 thumb screw or an integrated latch).
3536 2. Connect one end of the 14-pin ribbon cable to the G-Sync connector on the
3537 NVIDIA Quadro graphics card, and the other end to the NVIDIA SDI output
3540 3. On Quadro FX 4500/5500 SDI, connect the SMA-to-BNC cables by screwing the
3541 male SMA connectors onto the female SMA connectors on the NVIDIA SDI
3542 output card. On Quadro FX 4600/5600 SDI II, this step is not necessary:
3543 the SDI II has BNC connectors rather than SMA connectors.
3545 4. Connect the DVI-loopback connector by connecting one end of the DVI cable
3546 to the DVI connector on the NVIDIA SDI output card and the other end to
3547 the "north" DVI connector on the NVIDIA Quadro graphics card. The "north"
3548 DVI connector on the NVIDIA Quadro graphics card is the DVI connector
3549 that is the farthest from the graphics card PCI-E connection to the
3550 motherboard. The SDI output card will NOT function properly if this cable
3551 is connected to the "south" DVI connector.
3553 Once the above installation is complete, you may boot the system and configure
3554 the SDI video output using nvidia-settings. These instructions assume that you
3555 have already successfully installed the NVIDIA FreeBSD Accelerated Graphics
3556 Driver. If you have not done so, see Chapter 2 for details.
3559 22B. CLONE MODE CONFIGURATION WITH 'nvidia-settings'
3561 SDI video output is configured through the nvidia-settings utility. See the
3562 'nvidia-settings(1)' man page, and the nvidia-settings online help (click the
3563 "Help" button in the lower right corner of the interface for per-page help
3566 After the system has booted and X Windows has been started, run
3571 When the NVIDIA X Server Settings page appears, follow the steps below to
3572 configure the SDI video output.
3574 1. Click on the "Graphics to Video Out" tree item on the side menu. This
3575 will open the "Graphics to Video Out" page.
3577 2. Go to the "Synchronization Options" subpage and choose a synchronization
3578 method. From the "Sync Options" drop down click the list arrow to the
3579 right and then click the method that you want to use to synchronize the
3582 Sync Method Description
3583 ------------- --------------------------------------------------
3584 Free Running The SDI output will be synchronized with the
3585 timing chosen from the SDI signal format list.
3586 Genlock SDI output will be synchronized with the external
3588 Frame Lock The SDI output will be synchronized with the
3589 timing chosen from the SDI signal format list. In
3590 this case, the list of available timings is
3591 limited to those timings that can be synchronized
3592 with the detected external sync signal.
3595 Note that on Quadro FX 4600/5600 SDI II, you must first choose the
3596 correct Sync Format before an incoming sync signal will be detected.
3598 3. From the top Graphics to Video Out page, choose the output video format
3599 that will control the video resolution, field rate, and SMPTE signaling
3600 standard for the outgoing video stream. From the "Clone Mode" drop down
3601 box, click the "Video Format" arrow and then click the signal format that
3602 you would like to use. Note that only those resolutions that are smaller
3603 or equal to the desktop resolution will be available. Also, this list is
3604 pruned according to the sync option selected. If genlock synchronization
3605 is chosen, the output video format is automatically set to match the
3606 incoming video sync format and this drop down list will be grayed out
3607 preventing you from choosing another format. If frame lock
3608 synchronization has been selected, then only those modes that are
3609 compatible with the detected sync signal will be available.
3611 4. Choose the output data format from the "Output Data Format" drop down
3614 5. Click the "Enable SDI Output" button to enable video output using the
3615 settings above. The status of the SDI output can be verified by examining
3616 the LED indicators in the "Graphics to SDI property" page banner.
3618 6. To subsequently stop SDI output, simply click on the button that now says
3619 "Disable SDI Output".
3621 7. In order to change any of the SDI output parameters such as the Output
3622 Video Format, Output Data Format as well as the Synchronization Delay, it
3623 is necessary to first disable the SDI output.
3627 22C. CONFIGURATION FOR TWINVIEW OR AS A SEPARATE X SCREEN
3629 SDI video output can be configured through the nvidia-settings X Server
3630 Display Configuration page, for use in TwinView or as a separate X screen. The
3631 SDI video output can be configured as if it were a digital flat panel,
3632 choosing the resolution, refresh rate, and position within the desktop.
3634 Similarly, the SDI video output can be configured for use in TwinView or as a
3635 separate X screen through the X configuration file. The supported SDI video
3636 output modes can be requested by name anywhere a mode name can be used in the
3637 X configuration file (either in the "Modes" line, or in the "MetaModes"
3641 Option "MetaModes" "CRT-0:nvidia-auto-select, DFP-1:1280x720_60.00_smpte296"
3644 The mode names are reported in the nvidia-settings Display Configuration page
3645 when in advanced mode.
3647 Note that SDI "Clone Mode" as configured through the Graphics to Video Out
3648 page in nvidia-settings is mutually exclusive with using the SDI video output
3649 in TwinView or as a separate X screen.
3651 ______________________________________________________________________________
3653 Chapter 23. Configuring Depth 30 Displays
3654 ______________________________________________________________________________
3656 This driver release supports X screens with screen depths of 30 bits per pixel
3657 (10 bits per color component) on NVIDIA Quadro GPUs based on G80 and higher
3658 chip architectures. This provides about 1 billion possible colors, allowing
3659 for higher color precision and smoother gradients.
3661 When displaying a depth 30 image on a digital flat panel, the color data will
3662 be dithered to 8 or 6 bits per pixel, depending on the capabilities of the
3663 flat panel. VGA outputs can display the full 10 bit range of colors.
3665 To work reliably, depth 30 requires X.Org 7.3 or higher and pixman 0.11.6 or
3668 In addition to the above software requirements, many X applications and
3669 toolkits do not understand depth 30 visuals as of this writing. Some programs
3670 may work correctly, some may work but display incorrect colors, and some may
3671 simply fail to run. In particular, many OpenGL applications request 8 bits of
3672 alpha when searching for FBConfigs. Since depth 30 visuals have only 2 bits of
3673 alpha, no suitable FBConfigs will be found and such applications will fail to
3676 ______________________________________________________________________________
3678 Chapter 24. NVIDIA Contact Info and Additional Resources
3679 ______________________________________________________________________________
3681 If you believe that you have found a bug or have a problem that you need
3682 assistance with and cannot find the solution elsewhere, or if you have found
3683 inaccuracies in this document, send email to freebsd-gfx-bugs@nvidia.com
3687 Additional Resources
3689 XFree86 Video Timings HOWTO
3691 http://www.tldp.org/HOWTO/XFree86-Video-Timings-HOWTO/index.html
3693 The X.Org Foundation
3699 http://www.opengl.org/
3702 ______________________________________________________________________________
3705 ______________________________________________________________________________
3707 The port of the NVIDIA driver to FreeBSD is due in no small part to the many
3708 contributions of Christian Zander <zander 'at' mail.minion.de> and Matthew N.
3709 Dodd <mdodd 'at' freebsd.org>.
3711 ______________________________________________________________________________
3713 Chapter 26. Acknowledgements
3714 ______________________________________________________________________________
3716 The driver splash screen is decoded using 'libpng':
3717 http://libpng.org/pub/png/libpng.html
3719 This NVIDIA FreeBSD driver contains code from the int10 module of the X.Org
3722 The BSD implementations of the following compiler intrinsics are used for
3723 better portability: __udivdi3, __umoddi3, __divdi3, __moddi3, __ucmpdi2,
3724 __cmpdi2, __fixunssfdi, __fixunsdfdi, __ashldi3 and __lshrdi3.
3726 ______________________________________________________________________________
3728 Appendix E. Supported NVIDIA GPU Products
3729 ______________________________________________________________________________
3731 For the most complete and accurate listing of supported GPUs, please see the
3732 Supported Products List, available from the NVIDIA FreeBSD x86 Graphics Driver
3733 download page. Please go to http://www.nvidia.com/object/unix.html, follow the
3734 Archive link under the FreeBSD x86 heading, follow the link for the 180.29
3735 driver, and then go to the Supported Products List.
3738 E1. NVIDIA GEFORCE GPUS
3741 NVIDIA GPU product Device PCI ID
3742 ------------------------------------------------------ ---------------
3743 GeForce 6800 Ultra 0x0040
3745 GeForce 6800 LE 0x0042
3746 GeForce 6800 XE 0x0043
3747 GeForce 6800 XT 0x0044
3748 GeForce 6800 GT 0x0045
3749 GeForce 6800 GT 0x0046
3750 GeForce 6800 GS 0x0047
3751 GeForce 6800 XT 0x0048
3752 GeForce 7800 GTX 0x0090
3753 GeForce 7800 GTX 0x0091
3754 GeForce 7800 GT 0x0092
3755 GeForce 7800 GS 0x0093
3756 GeForce 7800 SLI 0x0095
3757 GeForce Go 7800 0x0098
3758 GeForce Go 7800 GTX 0x0099
3759 GeForce 6800 GS 0x00C0
3761 GeForce 6800 LE 0x00C2
3762 GeForce 6800 XT 0x00C3
3763 GeForce Go 6800 0x00C8
3764 GeForce Go 6800 Ultra 0x00C9
3766 GeForce 6600 GT 0x00F1
3769 GeForce 6600 LE 0x00F4
3770 GeForce 7800 GS 0x00F5
3771 GeForce 6800 GS 0x00F6
3772 GeForce 6800 Ultra 0x00F9
3773 GeForce 6600 GT 0x0140
3775 GeForce 6600 LE 0x0142
3776 GeForce 6600 VE 0x0143
3777 GeForce Go 6600 0x0144
3778 GeForce 6610 XL 0x0145
3779 GeForce Go 6600 TE/6200 TE 0x0146
3780 GeForce 6700 XL 0x0147
3781 GeForce Go 6600 0x0148
3782 GeForce Go 6600 GT 0x0149
3785 GeForce 6200 TurboCache(TM) 0x0161
3786 GeForce 6200SE TurboCache(TM) 0x0162
3787 GeForce 6200 LE 0x0163
3788 GeForce Go 6200 0x0164
3789 GeForce Go 6400 0x0166
3790 GeForce Go 6200 0x0167
3791 GeForce Go 6400 0x0168
3793 GeForce 7100 GS 0x016A
3794 GeForce 8800 GTX 0x0191
3795 GeForce 8800 GTS 0x0193
3796 GeForce 8800 Ultra 0x0194
3798 GeForce 7350 LE 0x01D0
3799 GeForce 7300 LE 0x01D1
3800 GeForce 7300 SE/7200 GS 0x01D3
3801 GeForce Go 7200 0x01D6
3802 GeForce Go 7300 0x01D7
3803 GeForce Go 7400 0x01D8
3804 GeForce 7500 LE 0x01DD
3805 GeForce 7300 GS 0x01DF
3807 GeForce 6800 LE 0x0212
3808 GeForce 6800 GT 0x0215
3809 GeForce 6800 XT 0x0218
3811 GeForce 6200 A-LE 0x0222
3813 GeForce 6150 LE 0x0241
3815 GeForce Go 6150 0x0244
3816 GeForce Go 6100 0x0247
3817 GeForce 7900 GTX 0x0290
3818 GeForce 7900 GT/GTO 0x0291
3819 GeForce 7900 GS 0x0292
3820 GeForce 7950 GX2 0x0293
3821 GeForce 7950 GX2 0x0294
3822 GeForce 7950 GT 0x0295
3823 GeForce Go 7950 GTX 0x0297
3824 GeForce Go 7900 GS 0x0298
3825 GeForce Go 7900 GTX 0x0299
3826 GeForce 7600 GT 0x02E0
3827 GeForce 7600 GS 0x02E1
3828 GeForce 7300 GT 0x02E2
3829 GeForce 7900 GS 0x02E3
3830 GeForce 7950 GT 0x02E4
3831 GeForce 7650 GS 0x0390
3832 GeForce 7600 GT 0x0391
3833 GeForce 7600 GS 0x0392
3834 GeForce 7300 GT 0x0393
3835 GeForce 7600 LE 0x0394
3836 GeForce 7300 GT 0x0395
3837 GeForce Go 7700 0x0397
3838 GeForce Go 7600 0x0398
3839 GeForce Go 7600 GT 0x0399
3840 GeForce 6150SE nForce 430 0x03D0
3841 GeForce 6100 nForce 405 0x03D1
3842 GeForce 6100 nForce 400 0x03D2
3843 GeForce 6100 nForce 420 0x03D5
3844 GeForce 8600 GTS 0x0400
3845 GeForce 8600 GT 0x0401
3846 GeForce 8600 GT 0x0402
3847 GeForce 8600GS 0x0403
3848 GeForce 8400 GS 0x0404
3849 GeForce 9500M GS 0x0405
3850 GeForce 8600M GT 0x0407
3851 GeForce 9650M GS 0x0408
3852 GeForce 8700M GT 0x0409
3853 GeForce 8400 SE 0x0420
3854 GeForce 8500 GT 0x0421
3855 GeForce 8400 GS 0x0422
3856 GeForce 8300 GS 0x0423
3857 GeForce 8400 GS 0x0424
3858 GeForce 8600M GS 0x0425
3859 GeForce 8400M GT 0x0426
3860 GeForce 8400M GS 0x0427
3861 GeForce 8400M G 0x0428
3862 GeForce 9400 GT 0x042C
3863 GeForce 9300M G 0x042E
3864 GeForce 7150M / nForce 630M 0x0531
3865 GeForce 7000M / nForce 610M 0x0533
3866 GeForce 7050 PV / NVIDIA nForce 630a 0x053A
3867 GeForce 7050 PV / NVIDIA nForce 630a 0x053B
3868 GeForce 7025 / NVIDIA nForce 630a 0x053E
3869 GeForce GTX 295 0x05E0
3870 GeForce GTX 280 0x05E1
3871 GeForce GTX 260 0x05E2
3872 GeForce GTX 285 0x05E3
3873 GeForce 8800 GTS 512 0x0600
3874 GeForce 9800 GT 0x0601
3875 GeForce 8800 GT 0x0602
3876 GeForce 9800 GX2 0x0604
3877 GeForce 9800 GT 0x0605
3878 GeForce 8800 GS 0x0606
3879 GeForce 9800M GTX 0x0608
3880 GeForce 8800M GTS 0x0609
3881 GeForce 9800M GT 0x060B
3882 GeForce 8800M GTX 0x060C
3883 GeForce 8800 GS 0x060D
3884 GeForce 9600 GSO 0x0610
3885 GeForce 8800 GT 0x0611
3886 GeForce 9800 GTX/9800 GTX+ 0x0612
3887 GeForce 9800 GTX+ 0x0613
3888 GeForce 9800 GT 0x0614
3889 GeForce 9800M GTX 0x0617
3890 GeForce 9600 GT 0x0622
3891 GeForce 9600 GS 0x0623
3892 GeForce 9800M GTS 0x0628
3893 GeForce 9700M GTS 0x062A
3894 GeForce 9800M GS 0x062B
3895 GeForce 9800M GTS 0x062C
3896 GeForce 9500 GT 0x0640
3897 GeForce 9500 GT 0x0643
3898 GeForce 9600M GT 0x0647
3899 GeForce 9600M GS 0x0648
3900 GeForce 9600M GT 0x0649
3901 GeForce 9700M GT 0x064A
3902 GeForce 9500M G 0x064B
3903 GeForce 9650M GT 0x064C
3904 GeForce 9650 S 0x0656
3905 GeForce 9300 GE 0x06E0
3906 GeForce 9300 GS 0x06E1
3907 GeForce 8400 GS 0x06E4
3908 GeForce 9300M GS 0x06E5
3909 GeForce 9200M GS 0x06E8
3910 GeForce 9300M GS 0x06E9
3911 GeForce 7150 / NVIDIA nForce 630i 0x07E0
3912 GeForce 7100 / NVIDIA nForce 630i 0x07E1
3913 GeForce 7050 / NVIDIA nForce 610i 0x07E3
3914 GeForce 9100M G 0x0844
3915 GeForce 8200M G 0x0845
3921 nForce 780a SLI 0x084C
3922 nForce 750a SLI 0x084D
3923 GeForce 8100 / nForce 720a 0x084F
3924 GeForce 9400M G 0x0862
3925 GeForce 9400M 0x0863
3929 E2. NVIDIA QUADRO GPUS
3932 NVIDIA GPU product Device PCI ID
3933 ------------------------------------------------------ ---------------
3934 Quadro FX 4000 0x004E
3935 Quadro FX 4500 0x009D
3936 Quadro FX Go1400 0x00CC
3937 Quadro FX 3450/4000 SDI 0x00CD
3938 Quadro FX 1400 0x00CE
3939 Quadro FX 4400/Quadro FX 3400 0x00F8
3940 Quadro NVS 440 0x014A
3941 Quadro FX 540M 0x014C
3942 Quadro FX 550 0x014D
3943 Quadro FX 540 0x014E
3944 Quadro NVS 285 0x0165
3945 Quadro FX 5600 0x019D
3946 Quadro FX 4600 0x019E
3947 Quadro NVS 110M 0x01D7
3948 Quadro NVS 110M 0x01DA
3949 Quadro NVS 120M 0x01DB
3950 Quadro FX 350M 0x01DC
3951 Quadro FX 350 0x01DE
3952 Quadro NVS 210S / NVIDIA GeForce 6150LE 0x0245
3953 Quadro FX 2500M 0x029A
3954 Quadro FX 1500M 0x029B
3955 Quadro FX 5500 0x029C
3956 Quadro FX 3500 0x029D
3957 Quadro FX 1500 0x029E
3958 Quadro FX 4500 X2 0x029F
3959 Quadro FX 560 0x039E
3960 Quadro FX 370 0x040A
3961 Quadro NVS 320M 0x040B
3962 Quadro FX 570M 0x040C
3963 Quadro FX 1600M 0x040D
3964 Quadro FX 570 0x040E
3965 Quadro FX 1700 0x040F
3966 Quadro NVS 140M 0x0429
3967 Quadro NVS 130M 0x042A
3968 Quadro NVS 135M 0x042B
3969 Quadro FX 360M 0x042D
3970 Quadro NVS 290 0x042F
3972 Quadro FX 5800 0x05FD
3973 Quadro FX 4800 0x05FE
3974 Quadro FX 3700 0x061A
3975 Quadro FX 3600M 0x061C
3976 Quadro FX 2700M 0x063A
3977 Quadro FX 770M 0x065C
3978 Quadro NVS 150M 0x06EA
3979 Quadro NVS 160M 0x06EB
3980 Quadro NVS 420 0x06F8
3981 Quadro FX 370 LP 0x06F9
3982 Quadro NVS 450 0x06FA
3983 Quadro NVS 295 0x06FD
3984 Quadro FX 470 0x087A
3985 Quadro FX 470M 0x087F
3988 Below are the legacy GPUs that are no longer supported in the unified driver.
3989 These GPUs will continue to be maintained through the special legacy NVIDIA
3990 GPU driver releases.
3992 The 173.14.xx driver supports the following set of GPUs:
3995 NVIDIA GPU product Device PCI ID
3996 ---------------------------------- ----------------------------------
3997 GeForce PCX 5750 0x00FA
3998 GeForce PCX 5900 0x00FB
3999 Quadro FX 330/GeForce PCX 5300 0x00FC
4000 Quadro FX 330/Quadro NVS 280 PCI-E 0x00FD
4001 Quadro FX 1300 0x00FE
4002 GeForce FX 5800 Ultra 0x0301
4003 GeForce FX 5800 0x0302
4004 Quadro FX 2000 0x0308
4005 Quadro FX 1000 0x0309
4006 GeForce FX 5600 Ultra 0x0311
4007 GeForce FX 5600 0x0312
4008 GeForce FX 5600XT 0x0314
4009 GeForce FX Go5600 0x031A
4010 GeForce FX Go5650 0x031B
4011 Quadro FX Go700 0x031C
4012 GeForce FX 5200 0x0320
4013 GeForce FX 5200 Ultra 0x0321
4014 GeForce FX 5200 0x0322
4015 GeForce FX 5200LE 0x0323
4016 GeForce FX Go5200 0x0324
4017 GeForce FX Go5250 0x0325
4018 GeForce FX 5500 0x0326
4019 GeForce FX 5100 0x0327
4020 GeForce FX Go5200 32M/64M 0x0328
4021 Quadro NVS 55/280 PCI 0x032A
4022 Quadro FX 500/FX 600 0x032B
4023 GeForce FX Go53xx 0x032C
4024 GeForce FX Go5100 0x032D
4025 GeForce FX 5900 Ultra 0x0330
4026 GeForce FX 5900 0x0331
4027 GeForce FX 5900XT 0x0332
4028 GeForce FX 5950 Ultra 0x0333
4029 GeForce FX 5900ZT 0x0334
4030 Quadro FX 3000 0x0338
4031 Quadro FX 700 0x033F
4032 GeForce FX 5700 Ultra 0x0341
4033 GeForce FX 5700 0x0342
4034 GeForce FX 5700LE 0x0343
4035 GeForce FX 5700VE 0x0344
4036 GeForce FX Go5700 0x0347
4037 GeForce FX Go5700 0x0348
4038 Quadro FX Go1000 0x034C
4039 Quadro FX 1100 0x034E
4042 The 96.43.xx driver supports the following set of GPUs:
4045 NVIDIA GPU product Device PCI ID
4046 ---------------------------------- ----------------------------------
4047 GeForce2 MX/MX 400 0x0110
4048 GeForce2 MX 100/200 0x0111
4050 Quadro2 MXR/EX/Go 0x0113
4051 GeForce4 MX 460 0x0170
4052 GeForce4 MX 440 0x0171
4053 GeForce4 MX 420 0x0172
4054 GeForce4 MX 440-SE 0x0173
4055 GeForce4 440 Go 0x0174
4056 GeForce4 420 Go 0x0175
4057 GeForce4 420 Go 32M 0x0176
4058 GeForce4 460 Go 0x0177
4059 Quadro4 550 XGL 0x0178
4060 GeForce4 440 Go 64M 0x0179
4061 Quadro NVS 400 0x017A
4062 Quadro4 500 GoGL 0x017C
4063 GeForce4 410 Go 16M 0x017D
4064 GeForce4 MX 440 with AGP8X 0x0181
4065 GeForce4 MX 440SE with AGP8X 0x0182
4066 GeForce4 MX 420 with AGP8X 0x0183
4067 GeForce4 MX 4000 0x0185
4068 Quadro4 580 XGL 0x0188
4069 Quadro NVS 280 SD 0x018A
4070 Quadro4 380 XGL 0x018B
4071 Quadro NVS 50 PCI 0x018C
4072 GeForce2 Integrated GPU 0x01A0
4073 GeForce4 MX Integrated GPU 0x01F0
4075 GeForce3 Ti 200 0x0201
4076 GeForce3 Ti 500 0x0202
4078 GeForce4 Ti 4600 0x0250
4079 GeForce4 Ti 4400 0x0251
4080 GeForce4 Ti 4200 0x0253
4081 Quadro4 900 XGL 0x0258
4082 Quadro4 750 XGL 0x0259
4083 Quadro4 700 XGL 0x025B
4084 GeForce4 Ti 4800 0x0280
4085 GeForce4 Ti 4200 with AGP8X 0x0281
4086 GeForce4 Ti 4800 SE 0x0282
4087 GeForce4 4200 Go 0x0286
4088 Quadro4 980 XGL 0x0288
4089 Quadro4 780 XGL 0x0289
4090 Quadro4 700 GoGL 0x028C
4093 The 71.86.xx driver supports the following set of GPUs:
4096 NVIDIA GPU product Device PCI ID
4097 ---------------------------------- ----------------------------------
4099 RIVA TNT2/TNT2 Pro 0x0028
4100 RIVA TNT2 Ultra 0x0029
4101 Vanta/Vanta LT 0x002C
4102 RIVA TNT2 Model 64/Model 64 Pro 0x002D
4107 GeForce2 GTS/GeForce2 Pro 0x0150
4109 GeForce2 Ultra 0x0152
4113 ______________________________________________________________________________
4115 Appendix F. X Config Options
4116 ______________________________________________________________________________
4118 The following driver options are supported by the NVIDIA X driver. They may be
4119 specified either in the Screen or Device sections of the X config file.
4123 Option "NvAGP" "integer"
4125 Configure AGP support. Integer argument can be one of:
4128 -------------- ---------------------------------------------------
4130 1 use NVIDIA internal AGP support, if possible
4131 2 use AGPGART, if possible
4132 3 use any AGP support (try AGPGART, then NVIDIA AGP)
4134 Note that NVIDIA internal AGP support cannot work if AGPGART is either
4135 statically compiled into your kernel or is built as a module and loaded
4136 into your kernel. See Chapter 9 for details. Default: 3.
4138 Option "NoLogo" "boolean"
4140 Disable drawing of the NVIDIA logo splash screen at X startup. Default:
4141 the logo is drawn for screens with depth 24.
4143 Option "LogoPath" "string"
4145 Sets the path to the PNG file to be used as the logo splash screen at X
4146 startup. If the PNG file specified has a bKGD (background color) chunk,
4147 then the screen is cleared to the color it specifies. Otherwise, the
4148 screen is cleared to black. The logo file must be owned by root and must
4149 not be writable by a non-root group. Note that a logo is only displayed
4150 for screens with depth 24. Default: The built-in NVIDIA logo is used.
4152 Option "RenderAccel" "boolean"
4154 Enable or disable hardware acceleration of the RENDER extension. Default:
4155 hardware acceleration of the RENDER extension is enabled.
4157 Option "NoRenderExtension" "boolean"
4159 Disable the RENDER extension. Other than recompiling it, the X server does
4160 not seem to have another way of disabling this. Fortunately, we can
4161 control this from the driver so we export this option. This is useful in
4162 depth 8 where RENDER would normally steal most of the default colormap.
4163 Default: RENDER is offered when possible.
4165 Option "UBB" "boolean"
4167 Enable or disable the Unified Back Buffer on Quadro-based GPUs (Quadro4
4168 NVS excluded); see Chapter 17 for a description of UBB. This option has no
4169 effect on non-Quadro GPU products. Default: UBB is on for Quadro GPUs.
4171 Option "NoFlip" "boolean"
4173 Disable OpenGL flipping; see Chapter 17 for a description. Default: OpenGL
4174 will swap by flipping when possible.
4176 Option "Dac8Bit" "boolean"
4178 Most Quadro products by default use a 10-bit color look-up table (LUT);
4179 setting this option to TRUE forces these GPUs to use an 8-bit (LUT).
4180 Default: a 10-bit LUT is used, when available.
4182 Option "Overlay" "boolean"
4184 Enables RGB workstation overlay visuals. This is only supported on Quadro
4185 GPUs (Quadro NVS GPUs excluded) in depth 24. This option causes the server
4186 to advertise the SERVER_OVERLAY_VISUALS root window property and GLX will
4187 report single- and double-buffered, Z-buffered 16-bit overlay visuals. The
4188 transparency key is pixel 0x0000 (hex). There is no gamma correction
4189 support in the overlay plane. This feature requires XFree86 version 4.2.0
4190 or newer, or the X.Org X server. When the X screen is either wider than
4191 2046 pixels or taller than 2047, the overlay may be emulated with a
4192 substantial performance penalty. RGB workstation overlays are not
4193 supported when the Composite extension is enabled.
4195 UBB must be enabled when overlays are enabled (this is the default
4198 Option "CIOverlay" "boolean"
4200 Enables Color Index workstation overlay visuals with identical
4201 restrictions to Option "Overlay" above. The server will offer visuals both
4202 with and without a transparency key. These are depth 8 PseudoColor
4203 visuals. Enabling Color Index overlays on X servers older than XFree86 4.3
4204 will force the RENDER extension to be disabled due to bugs in the RENDER
4205 extension in older X servers. Color Index workstation overlays are not
4206 supported when the Composite extension is enabled. Default: off.
4208 UBB must be enabled when overlays are enabled (this is the default
4211 Option "TransparentIndex" "integer"
4213 When color index overlays are enabled, use this option to choose which
4214 pixel is used for the transparent pixel in visuals featuring transparent
4215 pixels. This value is clamped between 0 and 255 (Note: some applications
4216 such as Alias's Maya require this to be zero in order to work correctly).
4219 Option "OverlayDefaultVisual" "boolean"
4221 When overlays are used, this option sets the default visual to an overlay
4222 visual thereby putting the root window in the overlay. This option is not
4223 recommended for RGB overlays. Default: off.
4225 Option "EmulatedOverlaysTimerMs" "integer"
4227 Enables the use of a timer within the X server to perform the updates to
4228 the emulated overlay or CI overlay. This option can be used to improve the
4229 performance of the emulated or CI overlays by reducing the frequency of
4230 the updates. The value specified indicates the desired number of
4231 milliseconds between overlay updates. To disable the use of the timer
4232 either leave the option unset or set it to 0. Default: off.
4234 Option "EmulatedOverlaysThreshold" "boolean"
4236 Enables the use of a threshold within the X server to perform the updates
4237 to the emulated overlay or CI overlay. The emulated or CI overlay updates
4238 can be deferred but this threshold will limit the number of deferred
4239 OpenGL updates allowed before the overlay is updated. This option can be
4240 used to trade off performance and animation quality. Default: on.
4242 Option "EmulatedOverlaysThresholdValue" "integer"
4244 Controls the threshold used in updating the emulated or CI overlays. This
4245 is used in conjunction with the EmulatedOverlaysThreshold option to trade
4246 off performance and animation quality. Higher values for this option favor
4247 performance over quality. Setting low values of this option will not cause
4248 the overlay to be updated more often than the frequence specified by the
4249 EmulatedOverlaysTimerMs option. Default: 5.
4251 Option "RandRRotation" "boolean"
4253 Enable rotation support for the XRandR extension. This allows use of the
4254 XRandR X server extension for configuring the screen orientation through
4255 rotation. This feature is supported using depth 24. This requires an X.Org
4256 X 6.8.1 or newer X server. This feature does not work with hardware
4257 overlays; emulated overlays will be used instead at a substantial
4258 performance penalty. See Chapter 14 for details. Default: off.
4260 Option "Rotate" "string"
4262 Enable static rotation support. Unlike the RandRRotation option above,
4263 this option takes effect as soon as the X server is started and will work
4264 with older versions of X. This feature is supported using depth 24. This
4265 feature does not work with hardware overlays; emulated overlays will be
4266 used instead at a substantial performance penalty. This option is not
4267 compatible with the RandR extension. Valid rotations are "normal", "left",
4268 "inverted", and "right". Default: off.
4270 Option "AllowDDCCI" "boolean"
4272 Enables DDC/CI support in the NV-CONTROL X extension. DDC/CI is a
4273 mechanism for communication between your computer and your display device.
4274 This can be used to set the values normally controlled through your
4275 display device's On Screen Display. See the DDC/CI NV-CONTROL attributes
4276 in 'NVCtrl.h' and functions in 'NVCtrlLib.h' in the 'nvidia-settings'
4277 source code. Default: off (DDC/CI is disabled).
4279 Note that support for DDC/CI within the NVIDIA X driver's NV-CONTROL
4280 extension is deprecated, and will be removed in a future release. Other
4281 mechanisms for DDC/CI, such as the kernel i2c subsystem on Linux, are
4282 preferred over NV-CONTROL's DDC/CI support.
4284 If you would prefer that the NVIDIA X driver's NV-CONTROL X extension not
4285 remove DDC/CI support, please make your concerns known my emailing
4286 linux-bugs@nvidia.com.
4288 Option "SWCursor" "boolean"
4290 Enable or disable software rendering of the X cursor. Default: off.
4292 Option "HWCursor" "boolean"
4294 Enable or disable hardware rendering of the X cursor. Default: on.
4296 Option "CursorShadow" "boolean"
4298 Enable or disable use of a shadow with the hardware accelerated cursor;
4299 this is a black translucent replica of your cursor shape at a given offset
4300 from the real cursor. Default: off (no cursor shadow).
4302 Option "CursorShadowAlpha" "integer"
4304 The alpha value to use for the cursor shadow; only applicable if
4305 CursorShadow is enabled. This value must be in the range [0, 255] -- 0 is
4306 completely transparent; 255 is completely opaque. Default: 64.
4308 Option "CursorShadowXOffset" "integer"
4310 The offset, in pixels, that the shadow image will be shifted to the right
4311 from the real cursor image; only applicable if CursorShadow is enabled.
4312 This value must be in the range [0, 32]. Default: 4.
4314 Option "CursorShadowYOffset" "integer"
4316 The offset, in pixels, that the shadow image will be shifted down from the
4317 real cursor image; only applicable if CursorShadow is enabled. This value
4318 must be in the range [0, 32]. Default: 2.
4320 Option "ConnectedMonitor" "string"
4322 Allows you to override what the NVIDIA kernel module detects is connected
4323 to your graphics card. This may be useful, for example, if you use a KVM
4324 (keyboard, video, mouse) switch and you are switched away when X is
4325 started. In such a situation, the NVIDIA kernel module cannot detect which
4326 display devices are connected, and the NVIDIA X driver assumes you have a
4329 Valid values for this option are "CRT" (cathode ray tube), "DFP" (digital
4330 flat panel), or "TV" (television); if using TwinView, this option may be a
4331 comma-separated list of display devices; e.g.: "CRT, CRT" or "CRT, DFP".
4333 It is generally recommended to not use this option, but instead use the
4334 "UseDisplayDevice" option.
4336 NOTE: anything attached to a 15 pin VGA connector is regarded by the
4337 driver as a CRT. "DFP" should only be used to refer to digital flat panels
4338 connected via a DVI port.
4340 Default: string is NULL (the NVIDIA driver will detect the connected
4343 Option "UseDisplayDevice" "string"
4345 The "UseDisplayDevice" X configuration option is a list of one or more
4346 display devices, which limits the display devices the NVIDIA X driver will
4347 consider for an X screen. The display device names used in the option may
4348 be either specific (with a numeric suffix; e.g., "DFP-1") or general
4349 (without a numeric suffix; e.g., "DFP").
4351 When assigning display devices to X screens, the NVIDIA X driver walks
4352 through the list of all (not already assigned) display devices detected as
4353 connected. When the "UseDisplayDevice" X configuration option is
4354 specified, the X driver will only consider connected display devices which
4355 are also included in the "UseDisplayDevice" list. This can be thought of
4356 as a "mask" against the connected (and not already assigned) display
4359 Note the subtle difference between this option and the "ConnectedMonitor"
4360 option: the "ConnectedMonitor" option overrides which display devices are
4361 actually detected, while the "UseDisplayDevice" option controls which of
4362 the detected display devices will be used on this X screen.
4364 Of the list of display devices considered for this X screen (either all
4365 connected display devices, or a subset limited by the "UseDisplayDevice"
4366 option), the NVIDIA X driver first looks at CRTs, then at DFPs, and
4367 finally at TVs. For example, if both a CRT and a DFP are connected, by
4368 default the X driver would assign the CRT to this X screen. However, by
4371 Option "UseDisplayDevice" "DFP"
4373 the X screen would use the DFP instead. Or, if CRT-0, DFP-0, and DFP-1 are
4374 connected and TwinView is enabled, the X driver would assign CRT-0 and
4375 DFP-0 to the X screen. However, by specifying:
4377 Option "UseDisplayDevice" "CRT-0, DFP-1"
4379 the X screen would use CRT-0 and DFP-1 instead.
4381 Additionally, the special value "none" can be specified for the
4382 "UseDisplayDevice" option. When this value is given, any programming of
4383 the display hardware is disabled. The NVIDIA driver will not perform any
4384 mode validation or mode setting for this X screen. This is intended for
4385 use in conjunction with CUDA or in remote graphics solutions such as VNC
4386 or Hewlett Packard's Remote Graphics Software (RGS). This functionality is
4387 only available on Quadro and Tesla GPUs.
4389 Note the following restrictions for setting the "UseDisplayDevice" to
4392 o OpenGL SyncToVBlank will have no effect.
4394 o None of Stereo, Overlay, CIOverlay, or SLI are allowed when
4395 "UseDisplayDevice" is set to "none".
4398 Option "UseEdidFreqs" "boolean"
4400 This option controls whether the NVIDIA X driver will use the HorizSync
4401 and VertRefresh ranges given in a display device's EDID, if any. When
4402 UseEdidFreqs is set to True, EDID-provided range information will override
4403 the HorizSync and VertRefresh ranges specified in the Monitor section. If
4404 a display device does not provide an EDID, or the EDID does not specify an
4405 hsync or vrefresh range, then the X server will default to the HorizSync
4406 and VertRefresh ranges specified in the Monitor section of your X config
4407 file. These frequency ranges are used when validating modes for your
4410 Default: True (EDID frequencies will be used)
4412 Option "UseEDID" "boolean"
4414 By default, the NVIDIA X driver makes use of a display device's EDID, when
4415 available, during construction of its mode pool. The EDID is used as a
4416 source for possible modes, for valid frequency ranges, and for collecting
4417 data on the physical dimensions of the display device for computing the
4418 DPI (see Appendix I). However, if you wish to disable the driver's use of
4419 the EDID, you can set this option to False:
4421 Option "UseEDID" "FALSE"
4423 Note that, rather than globally disable all uses of the EDID, you can
4424 individually disable each particular use of the EDID; e.g.,
4426 Option "UseEDIDFreqs" "FALSE"
4427 Option "UseEDIDDpi" "FALSE"
4428 Option "ModeValidation" "NoEdidModes"
4430 Default: True (use EDID).
4432 Option "IgnoreEDID" "boolean"
4434 This option is deprecated, and no longer affects behavior of the X driver.
4435 See the "UseEDID" option for details.
4437 Option "NoDDC" "boolean"
4439 Synonym for "IgnoreEDID". This option is deprecated, and no longer affects
4440 behavior of the X driver. See the "UseEDID" option for details.
4442 Option "UseInt10Module" "boolean"
4444 Enable use of the X Int10 module to soft-boot all secondary cards, rather
4445 than POSTing the cards through the NVIDIA kernel module. Default: off
4446 (POSTing is done through the NVIDIA kernel module).
4448 Option "TwinView" "boolean"
4450 Enable or disable TwinView. See Chapter 10 for details. Default: off
4451 (TwinView is disabled).
4453 Option "TwinViewOrientation" "string"
4455 Controls the relationship between the two display devices when using
4456 TwinView. Takes one of the following values: "RightOf" "LeftOf" "Above"
4457 "Below" "Clone". See Chapter 10 for details. Default: string is NULL.
4459 Option "SecondMonitorHorizSync" "range(s)"
4461 This option is like the HorizSync entry in the Monitor section, but is for
4462 the second monitor when using TwinView. See Chapter 10 for details.
4465 Option "SecondMonitorVertRefresh" "range(s)"
4467 This option is like the VertRefresh entry in the Monitor section, but is
4468 for the second monitor when using TwinView. See Chapter 10 for details.
4471 Option "MetaModes" "string"
4473 This option describes the combination of modes to use on each monitor when
4474 using TwinView. See Chapter 10 for details. Default: string is NULL.
4476 Option "NoTwinViewXineramaInfo" "boolean"
4478 When in TwinView, the NVIDIA X driver normally provides a Xinerama
4479 extension that X clients (such as window managers) can use to discover the
4480 current TwinView configuration, such as where each display device is
4481 positioned within the X screen. Some window mangers get confused by this
4482 information, so this option is provided to disable this behavior. Default:
4483 false (TwinView Xinerama information is provided).
4485 Due to bugs in some older software, TwinView Xinerama information is not
4486 provided by default on X.Org 7.1 and older when the X server is started
4487 with only one display device connected.
4489 Option "TwinViewXineramaInfoOrder" "string"
4491 When the NVIDIA X driver provides TwinViewXineramaInfo (see the
4492 NoTwinViewXineramaInfo X config option), it by default reports the
4493 currently enabled display devices in the order "CRT, DFP, TV". The
4494 TwinViewXineramaInfoOrder X config option can be used to override this
4497 The option string is a comma-separated list of display device names. The
4498 display device names can either be general (e.g, "CRT", which identifies
4499 all CRTs), or specific (e.g., "CRT-1", which identifies a particular CRT).
4500 Not all display devices need to be identified in the option string;
4501 display devices that are not listed will be implicitly appended to the end
4502 of the list, in their default order.
4504 Note that TwinViewXineramaInfoOrder tracks all display devices that could
4505 possibly be connected to the GPU, not just the ones that are currently
4506 enabled. When reporting the Xinerama information, the NVIDIA X driver
4507 walks through the display devices in the order specified, only reporting
4508 enabled display devices.
4514 "DFP-1, DFP-0, TV, CRT"
4516 In the first example, any enabled DFPs would be reported first (any
4517 enabled CRTs or TVs would be reported afterwards). In the second example,
4518 any enabled TVs would be reported first, then any enabled DFPs (any
4519 enabled CRTs would be reported last). In the last example, if DFP-1 were
4520 enabled, it would be reported first, then DFP-0, then any enabled TVs, and
4521 then any enabled CRTs; finally, any other enabled DFPs would be reported.
4523 Default: "CRT, DFP, TV"
4525 Option "TwinViewXineramaInfoOverride" "string"
4527 This option overrides the values reported by NVIDIA's TwinView Xinerama
4528 implementation. This disregards the actual display devices used by the X
4529 screen and any order specified in TwinViewXineramaInfoOrder.
4531 The option string is interpreted as a comma-separated list of regions,
4532 specified as '[width]x[height]+[x-offset]+[y-offset]'. The regions' sizes
4533 and offsets are not validated against the X screen size, but are directly
4534 reported to any Xinerama client.
4538 "1600x1200+0+0, 1600x1200+1600+0"
4539 "1024x768+0+0, 1024x768+1024+0, 1024x768+0+768, 1024x768+1024+768"
4542 Option "TVStandard" "string"
4544 See Chapter 13 for details on configuring TV-out.
4546 Option "TVOutFormat" "string"
4548 See Chapter 13 for details on configuring TV-out.
4550 Option "TVOverScan" "Decimal value in the range 0.0 to 1.0"
4552 Valid values are in the range 0.0 through 1.0; See Chapter 13 for details
4553 on configuring TV-out.
4555 Option "Stereo" "integer"
4557 Enable offering of quad-buffered stereo visuals on Quadro. Integer
4558 indicates the type of stereo equipment being used:
4561 -------------- ---------------------------------------------------
4562 1 DDC glasses. The sync signal is sent to the
4563 glasses via the DDC signal to the monitor. These
4564 usually involve a passthrough cable between the
4565 monitor and the graphics card. This mode is not
4566 available on G8xGL and higher GPUs.
4567 2 "Blueline" glasses. These usually involve a
4568 passthrough cable between the monitor and graphics
4569 card. The glasses know which eye to display based
4570 on the length of a blue line visible at the bottom
4571 of the screen. When in this mode, the root window
4572 dimensions are one pixel shorter in the Y
4573 dimension than requested. This mode does not work
4574 with virtual root window sizes larger than the
4575 visible root window size (desktop panning). This
4576 mode is not available on G8xGL and higher GPUs.
4577 3 Onboard stereo support. This is usually only found
4578 on professional cards. The glasses connect via a
4579 DIN connector on the back of the graphics card.
4580 4 TwinView clone mode stereo (also known as
4581 "passive" stereo). On graphics cards that support
4582 TwinView, the left eye is displayed on the first
4583 display, and the right eye is displayed on the
4584 second display. This is normally used in
4585 conjunction with special projectors to produce 2
4586 polarized images which are then viewed with
4587 polarized glasses. To use this stereo mode, you
4588 must also configure TwinView in clone mode with
4589 the same resolution, panning offset, and panning
4590 domains on each display.
4591 5 Vertical interlaced stereo mode, for use with
4592 SeeReal Stereo Digital Flat Panels.
4593 6 Color interleaved stereo mode, for use with
4594 Sharp3D Stereo Digital Flat Panels.
4596 Stereo is only available on Quadro cards. Stereo options 1, 2, and 3 (also
4597 known as "active" stereo) may be used with TwinView if all modes within
4598 each MetaMode have identical timing values. See Chapter 16 for suggestions
4599 on making sure the modes within your MetaModes are identical. The
4600 identical ModeLine requirement is not necessary for Stereo option 4
4601 ("passive" stereo). Default: 0 (Stereo is not enabled).
4603 UBB must be enabled when stereo is enabled (this is the default behavior).
4605 Stereo options 1, 2, and 3 ("active" stereo) are not supported on digital
4608 Multi-GPU cards (such as the Quadro FX 4500 X2) provide a single connector
4609 for onboard stereo support (option 3), which is tied to the bottommost
4610 GPU. In order to synchronize onboard stereo with the other GPU, you must
4611 use a G-Sync device (see Chapter 21 for details).
4613 Option "AllowDFPStereo" "boolean"
4615 By default, the NVIDIA X driver performs a check which disables active
4616 stereo (stereo options 1, 2, and 3) if the X screen is driving a DFP. The
4617 "AllowDFPStereo" option bypasses this check.
4619 Option "ForceStereoFlipping" "boolean"
4621 Stereo flipping is the process by which left and right eyes are displayed
4622 on alternating vertical refreshes. Normally, stereo flipping is only
4623 performed when a stereo drawable is visible. This option forces stereo
4624 flipping even when no stereo drawables are visible.
4626 This is to be used in conjunction with the "Stereo" option. If "Stereo" is
4627 0, the "ForceStereoFlipping" option has no effect. If otherwise, the
4628 "ForceStereoFlipping" option will force the behavior indicated by the
4629 "Stereo" option, even if no stereo drawables are visible. This option is
4630 useful in a multiple-screen environment in which a stereo application is
4631 run on a different screen than the stereo master.
4636 -------------- ---------------------------------------------------
4637 0 Stereo flipping is not forced. The default
4638 behavior as indicated by the "Stereo" option is
4640 1 Stereo flipping is forced. Stereo is running even
4641 if no stereo drawables are visible. The stereo
4642 mode depends on the value of the "Stereo" option.
4644 Default: 0 (Stereo flipping is not forced). Note that active stereo is not
4645 supported on digital flat panels.
4647 Option "XineramaStereoFlipping" "boolean"
4649 By default, when using Stereo with Xinerama, all physical X screens having
4650 a visible stereo drawable will stereo flip. Use this option to allow only
4651 one physical X screen to stereo flip at a time.
4653 This is to be used in conjunction with the "Stereo" and "Xinerama"
4654 options. If "Stereo" is 0 or "Xinerama" is 0, the "XineramaStereoFlipping"
4655 option has no effect.
4657 If you wish to have all X screens stereo flip all the time, see the
4658 "ForceStereoFlipping" option.
4663 -------------- ---------------------------------------------------
4664 0 Stereo flipping is enabled on one X screen at a
4665 time. Stereo is enabled on the first X screen
4666 having the stereo drawable.
4667 1 Stereo flipping in enabled on all X screens.
4669 Default: 1 (Stereo flipping is enabled on all X screens).
4671 Option "NoBandWidthTest" "boolean"
4673 As part of mode validation, the X driver tests if a given mode fits within
4674 the hardware's memory bandwidth constraints. This option disables this
4675 test. Default: false (the memory bandwidth test is performed).
4677 Option "IgnoreDisplayDevices" "string"
4679 This option tells the NVIDIA kernel module to completely ignore the
4680 indicated classes of display devices when checking which display devices
4681 are connected. You may specify a comma-separated list containing any of
4682 "CRT", "DFP", and "TV". For example:
4684 Option "IgnoreDisplayDevices" "DFP, TV"
4686 will cause the NVIDIA driver to not attempt to detect if any digital flat
4687 panels or TVs are connected. This option is not normally necessary;
4688 however, some video BIOSes contain incorrect information about which
4689 display devices may be connected, or which i2c port should be used for
4690 detection. These errors can cause long delays in starting X. If you are
4691 experiencing such delays, you may be able to avoid this by telling the
4692 NVIDIA driver to ignore display devices which you know are not connected.
4693 NOTE: anything attached to a 15 pin VGA connector is regarded by the
4694 driver as a CRT. "DFP" should only be used to refer to digital flat panels
4695 connected via a DVI port.
4697 Option "MultisampleCompatibility" "boolean"
4699 Enable or disable the use of separate front and back multisample buffers.
4700 Enabling this will consume more memory but is necessary for correct output
4701 when rendering to both the front and back buffers of a multisample or FSAA
4702 drawable. This option is necessary for correct operation of SoftImage XSI.
4703 Default: false (a single multisample buffer is shared between the front
4706 Option "NoPowerConnectorCheck" "boolean"
4708 The NVIDIA X driver will abort X server initialization if it detects that
4709 a GPU that requires an external power connector does not have an external
4710 power connector plugged in. This option can be used to bypass this test.
4711 Default: false (the power connector test is performed).
4713 Option "XvmcUsesTextures" "boolean"
4715 Forces XvMC to use the 3D engine for XvMCPutSurface requests rather than
4716 the video overlay. Default: false (video overlay is used when available).
4718 Option "AllowGLXWithComposite" "boolean"
4720 Enables GLX even when the Composite X extension is loaded. ENABLE AT YOUR
4721 OWN RISK. OpenGL applications will not display correctly in many
4722 circumstances with this setting enabled.
4724 This option is intended for use on X.Org X servers older than X11R6.9.0.
4725 On X11R6.9.0 or newer X servers, the NVIDIA OpenGL implementation
4726 interacts properly by default with the Composite X extension and this
4727 option should not be needed. However, on X11R6.9.0 or newer X servers,
4728 support for GLX with Composite can be disabled by setting this option to
4731 Default: false (GLX is disabled when Composite is enabled on X servers
4732 older than X11R6.9.0).
4734 Option "UseCompositeWrapper" "boolean"
4736 Enables the X server's "composite wrapper", which performs coordinate
4737 translations necessary for the Composite extension.
4739 Default: false (the NVIDIA X driver performs its own coordinate
4742 Option "AddARGBGLXVisuals" "boolean"
4744 Adds a 32-bit ARGB visual for each supported OpenGL configuration. This
4745 allows applications to use OpenGL to render with alpha transparency into
4746 32-bit windows and pixmaps. This option requires the Composite extension.
4747 Default: ARGB GLX visuals are enabled on X servers new enough to support
4748 them when the Composite extension is also enabled.
4750 Option "DisableGLXRootClipping" "boolean"
4752 If enabled, no clipping will be performed on rendering done by OpenGL in
4753 the root window. This option is deprecated. It is needed by older versions
4754 of OpenGL-based composite managers that draw the contents of redirected
4755 windows directly into the root window using OpenGL. Most OpenGL-based
4756 composite managers have been updated to support the Composite Overlay
4757 Window, a feature introduced in Xorg release 7.1. Using the Composite
4758 Overlay Window is the preferred method for performing OpenGL-based
4761 Option "DamageEvents" "boolean"
4763 Use OS-level events to efficiently notify X when a client has performed
4764 direct rendering to a window that needs to be composited. This will
4765 significantly improve performance and interactivity when using GLX
4766 applications with a composite manager running. It will also affect
4767 applications using GLX when rotation is enabled. This option is currently
4768 incompatible with SLI and Multi-GPU modes and will be disabled if either
4769 are used. Enabled by default.
4771 Option "ExactModeTimingsDVI" "boolean"
4773 Forces the initialization of the X server with the exact timings specified
4774 in the ModeLine. Default: false (for DVI devices, the X server initializes
4775 with the closest mode in the EDID list).
4777 Option "Coolbits" "integer"
4779 Enables various unsupported features, such as support for GPU clock
4780 manipulation in the NV-CONTROL X extension. This option accepts a bit mask
4781 of features to enable.
4783 When "1" (Bit 0) is set in the "Coolbits" option value, the
4784 nvidia-settings utility will contain a page labeled "Clock Frequencies"
4785 through which clock settings can be manipulated. "Coolbits" is only
4786 available on GeForce FX, Quadro FX and newer desktop GPUs. On GeForce FX
4787 and newer mobile GPUs, limited clock manipulation support is available
4788 when "1" is set in the "Coolbits" option value: clocks can be lowered
4789 relative to the default settings; overclocking is not supported due to the
4790 thermal constraints of notebook designs.
4792 WARNING: this may cause system damage and void warranties. This utility
4793 can run your computer system out of the manufacturer's design
4794 specifications, including, but not limited to: higher system voltages,
4795 above normal temperatures, excessive frequencies, and changes to BIOS that
4796 may corrupt the BIOS. Your computer's operating system may hang and result
4797 in data loss or corrupted images. Depending on the manufacturer of your
4798 computer system, the computer system, hardware and software warranties may
4799 be voided, and you may not receive any further manufacturer support.
4800 NVIDIA does not provide customer service support for the Coolbits option.
4801 It is for these reasons that absolutely no warranty or guarantee is either
4802 express or implied. Before enabling and using, you should determine the
4803 suitability of the utility for your intended use, and you shall assume all
4804 responsibility in connection therewith.
4806 When "2" (Bit 1) is set in the "Coolbits" option value, the NVIDIA driver
4807 will attempt to initialize SLI when using GPUs with different amounts of
4810 The default for this option is 0 (unsupported features are disabled).
4812 Option "MultiGPU" "string"
4814 This option controls the configuration of Multi-GPU rendering in supported
4818 -------------------------------- --------------------------------
4819 0, no, off, false, Single Use only a single GPU when
4821 1, yes, on, true, Auto Enable Multi-GPU and allow the
4822 driver to automatically select
4823 the appropriate rendering mode.
4824 AFR Enable Multi-GPU and use the
4825 Alternate Frame Rendering mode.
4826 SFR Enable Multi-GPU and use the
4827 Split Frame Rendering mode.
4828 AA Enable Multi-GPU and use
4829 antialiasing. Use this in
4830 conjunction with full scene
4831 antialiasing to improve visual
4835 Option "SLI" "string"
4837 This option controls the configuration of SLI rendering in supported
4841 -------------------------------- --------------------------------
4842 0, no, off, false, Single Use only a single GPU when
4844 1, yes, on, true, Auto Enable SLI and allow the driver
4845 to automatically select the
4846 appropriate rendering mode.
4847 AFR Enable SLI and use the Alternate
4848 Frame Rendering mode.
4849 SFR Enable SLI and use the Split
4850 Frame Rendering mode.
4851 AA Enable SLI and use SLI
4852 Antialiasing. Use this in
4853 conjunction with full scene
4854 antialiasing to improve visual
4856 AFRofAA Enable SLI and use SLI Alternate
4857 Frame Rendering of Antialiasing
4858 mode. Use this in conjunction
4859 with full scene antialiasing to
4860 improve visual quality. This
4861 option is only valid for SLI
4862 configurations with 4 GPUs.
4865 Option "TripleBuffer" "boolean"
4867 Enable or disable the use of triple buffering. If this option is enabled,
4868 OpenGL windows that sync to vblank and are double-buffered will be given a
4869 third buffer. This decreases the time an application stalls while waiting
4870 for vblank events, but increases latency slightly (delay between user
4871 input and displayed result).
4873 Option "DPI" "string"
4875 This option specifies the Dots Per Inch for the X screen; for example:
4877 Option "DPI" "75 x 85"
4879 will set the horizontal DPI to 75 and the vertical DPI to 85. By default,
4880 the X driver will compute the DPI of the X screen from the EDID of any
4881 connected display devices. See Appendix I for details. Default: string is
4884 Option "UseEdidDpi" "string"
4886 By default, the NVIDIA X driver computes the DPI of an X screen based on
4887 the physical size of the display device, as reported in the EDID, and the
4888 size in pixels of the first mode to be used on the display device. If
4889 multiple display devices are used by the X screen, then the NVIDIA X
4890 screen will choose which display device to use. This option can be used to
4891 specify which display device to use. The string argument can be a display
4892 device name, such as:
4894 Option "UseEdidDpi" "DFP-0"
4896 or the argument can be "FALSE" to disable use of EDID-based DPI
4899 Option "UseEdidDpi" "FALSE"
4901 See Appendix I for details. Default: string is NULL (the driver computes
4902 the DPI from the EDID of a display device and selects the display device).
4904 Option "ConstantDPI" "boolean"
4906 By default on X.Org 6.9 or newer X servers, the NVIDIA X driver recomputes
4907 the size in millimeters of the X screen whenever the size in pixels of the
4908 X screen is changed using XRandR, such that the DPI remains constant.
4910 This behavior can be disabled (which means that the size in millimeters
4911 will not change when the size in pixels of the X screen changes) by
4912 setting the "ConstantDPI" option to "FALSE"; e.g.,
4914 Option "ConstantDPI" "FALSE"
4916 ConstantDPI defaults to True.
4918 On X servers older than X.Org 6.9, the NVIDIA X driver cannot change the
4919 size in millimeters of the X screen. Therefore the DPI of the X screen
4920 will change when XRandR changes the size in pixels of the X screen. The
4921 driver will behave as if ConstantDPI was forced to FALSE.
4923 Option "CustomEDID" "string"
4925 This option forces the X driver to use the EDID specified in a file rather
4926 than the display's EDID. You may specify a semicolon separated list of
4927 display names and filename pairs. The display name is any of "CRT-0",
4928 "CRT-1", "DFP-0", "DFP-1", "TV-0", "TV-1", or one of the generic names
4929 "CRT", "DFP", "TV", which apply the EDID to all devices of the specified
4930 type. The file contains a raw EDID (e.g., a file generated by
4935 Option "CustomEDID" "CRT-0:/tmp/edid1.bin; DFP-0:/tmp/edid2.bin"
4937 will assign the EDID from the file /tmp/edid1.bin to the display device
4938 CRT-0, and the EDID from the file /tmp/edid2.bin to the display device
4939 DFP-0. Note that a display device name must always be specified even if
4940 only one EDID is specified.
4942 Caution: Specifying an EDID that doesn't exactly match your display may
4943 damage your hardware, as it allows the driver to specify timings beyond
4944 the capabilities of your display. Use with care.
4946 Option "ModeValidation" "string"
4948 This option provides fine-grained control over each stage of the mode
4949 validation pipeline, disabling individual mode validation checks. This
4950 option should only very rarely be used.
4952 The option string is a semicolon-separated list of comma-separated lists
4953 of mode validation arguments. Each list of mode validation arguments can
4954 optionally be prepended with a display device name.
4956 "<dpy-0>: <tok>, <tok>; <dpy-1>: <tok>, <tok>, <tok>; ..."
4961 o "AllowNon60HzDFPModes": some lower quality TMDS encoders are only
4962 rated to drive DFPs at 60Hz; the driver will determine when only 60Hz
4963 DFP modes are allowed. This argument disables this stage of the mode
4964 validation pipeline.
4966 o "NoMaxPClkCheck": each mode has a pixel clock; this pixel clock is
4967 validated against the maximum pixel clock of the hardware (for a DFP,
4968 this is the maximum pixel clock of the TMDS encoder, for a CRT, this
4969 is the maximum pixel clock of the DAC). This argument disables the
4970 maximum pixel clock checking stage of the mode validation pipeline.
4972 o "NoEdidMaxPClkCheck": a display device's EDID can specify the maximum
4973 pixel clock that the display device supports; a mode's pixel clock is
4974 validated against this pixel clock maximum. This argument disables
4975 this stage of the mode validation pipeline.
4977 o "AllowInterlacedModes": interlaced modes are not supported on all
4978 NVIDIA GPUs; the driver will discard interlaced modes on GPUs where
4979 interlaced modes are not supported; this argument disables this stage
4980 of the mode validation pipeline.
4982 o "NoMaxSizeCheck": each NVIDIA GPU has a maximum resolution that it
4983 can drive; this argument disables this stage of the mode validation
4986 o "NoHorizSyncCheck": a mode's horizontal sync is validated against the
4987 range of valid horizontal sync values; this argument disables this
4988 stage of the mode validation pipeline.
4990 o "NoVertRefreshCheck": a mode's vertical refresh rate is validated
4991 against the range of valid vertical refresh rate values; this
4992 argument disables this stage of the mode validation pipeline.
4994 o "NoWidthAlignmentCheck": the alignment of a mode's visible width is
4995 validated against the capabilities of the GPU; normally, a mode's
4996 visible width must be a multiple of 8. This argument disables this
4997 stage of the mode validation pipeline.
4999 o "NoDFPNativeResolutionCheck": when validating for a DFP, a mode's
5000 size is validated against the native resolution of the DFP; this
5001 argument disables this stage of the mode validation pipeline.
5003 o "NoVirtualSizeCheck": if the X configuration file requests a specific
5004 virtual screen size, a mode cannot be larger than that virtual size;
5005 this argument disables this stage of the mode validation pipeline.
5007 o "NoVesaModes": when constructing the mode pool for a display device,
5008 the X driver uses a built-in list of VESA modes as one of the mode
5009 sources; this argument disables use of these built-in VESA modes.
5011 o "NoEdidModes": when constructing the mode pool for a display device,
5012 the X driver uses any modes listed in the display device's EDID as
5013 one of the mode sources; this argument disables use of EDID-specified
5016 o "NoXServerModes": when constructing the mode pool for a display
5017 device, the X driver uses the built-in modes provided by the core
5018 XFree86/Xorg X server as one of the mode sources; this argument
5019 disables use of these modes. Note that this argument does not disable
5020 custom ModeLines specified in the X config file; see the
5021 "NoCustomModes" argument for that.
5023 o "NoCustomModes": when constructing the mode pool for a display
5024 device, the X driver uses custom ModeLines specified in the X config
5025 file (through the "Mode" or "ModeLine" entries in the Monitor
5026 Section) as one of the mode sources; this argument disables use of
5029 o "NoPredefinedModes": when constructing the mode pool for a display
5030 device, the X driver uses additional modes predefined by the NVIDIA X
5031 driver; this argument disables use of these modes.
5033 o "NoUserModes": additional modes can be added to the mode pool
5034 dynamically, using the NV-CONTROL X extension; this argument
5035 prohibits user-specified modes via the NV-CONTROL X extension.
5037 o "NoExtendedGpuCapabilitiesCheck": allow mode timings that may exceed
5038 the GPU's extended capability checks.
5040 o "ObeyEdidContradictions": an EDID may contradict itself by listing a
5041 mode as supported, but the mode may exceed an EDID-specified valid
5042 frequency range (HorizSync, VertRefresh, or maximum pixel clock).
5043 Normally, the NVIDIA X driver prints a warning in this scenario, but
5044 does not invalidate an EDID-specified mode just because it exceeds an
5045 EDID-specified valid frequency range. However, the
5046 "ObeyEdidContradictions" argument instructs the NVIDIA X driver to
5047 invalidate these modes.
5049 o "NoTotalSizeCheck": allow modes in which the individual visible or
5050 sync pulse timings exceed the total raster size.
5052 o "DoubleScanPriority": on GPUs older than G80, doublescan modes are
5053 sorted before non-doublescan modes of the same resolution for
5054 purposes of mode pool sorting; but on G80 and later GPUs, doublescan
5055 modes are sorted after non-doublescan modes of the same resolution.
5056 This token inverts that priority (i.e., doublescan modes will be
5057 sorted after on pre-G80 GPUs, and sorted before on G80 and later
5060 o "NoDualLinkDVICheck": for mode timings used on dual link DVI DFPs,
5061 the driver must perform additional checks to ensure that the correct
5062 pixels are sent on the correct link. For some of these checks, the
5063 driver will invalidate the mode timings; for other checks, the driver
5064 will implicitly modify the mode timings to meet the GPU's dual link
5065 DVI requirements. This token disables this dual link DVI checking.
5070 Option "ModeValidation" "NoMaxPClkCheck"
5072 disable the maximum pixel clock check when validating modes on all display
5075 Option "ModeValidation" "CRT-0: NoEdidModes, NoMaxPClkCheck; DFP-0:
5078 do not use EDID modes and do not perform the maximum pixel clock check on
5079 CRT-0, and do not use VESA modes on DFP-0.
5081 Option "ModeDebug" "boolean"
5083 This option causes the X driver to print verbose details about mode
5084 validation to the X log file. Note that this option is applied globally:
5085 setting this option to TRUE will enable verbose mode validation logging
5086 for all NVIDIA X screens in the X server.
5088 Option "UseEvents" "boolean"
5090 Enables the use of system events in some cases when the X driver is
5091 waiting for the hardware. The X driver can briefly spin through a tight
5092 loop when waiting for the hardware. With this option the X driver instead
5093 sets an event handler and waits for the hardware through the 'poll()'
5094 system call. Default: the use of the events is disabled.
5096 Option "FlatPanelProperties" "string"
5098 This option requests particular properties for all or a subset of the
5099 connected flat panels.
5101 The option string is a semicolon-separated list of comma-separated
5102 property=value pairs. Each list of property=value pairs can optionally be
5103 prepended with a flat panel name.
5105 "<DFP-0>: <property=value>, <property=value>; <DFP-1>:
5106 <property=value>; ..."
5109 Recognized properties:
5111 o "Scaling": controls the flat panel scaling mode; possible values are:
5112 'Default' (the driver will use whichever scaling state is current),
5113 'Native' (the driver will use the flat panel's scaler, if possible),
5114 'Scaled' (the driver will use the NVIDIA GPU's scaler, if possible),
5115 'Centered' (the driver will center the image, if possible), and
5116 'aspect-scaled' (the X driver will scale with the NVIDIA GPU's
5117 scaler, but keep the aspect ratio correct).
5119 o "Dithering": controls the flat panel dithering mode; possible values
5120 are: 'Default' (the driver will decide when to dither), 'Enabled'
5121 (the driver will always dither, if possible), and 'Disabled' (the
5122 driver will never dither).
5127 Option "FlatPanelProperties" "Scaling = Centered"
5129 set the flat panel scaling mode to centered on all flat panels.
5131 Option "FlatPanelProperties" "DFP-0: Scaling = Centered; DFP-1:
5132 Scaling = Scaled, Dithering = Enabled"
5134 set DFP-0's scaling mode to centered, set DFP-1's scaling mode to scaled
5135 and its dithering mode to enabled.
5137 Option "ProbeAllGpus" "boolean"
5139 When the NVIDIA X driver initializes, it probes all GPUs in the system,
5140 even if no X screens are configured on them. This is done so that the X
5141 driver can report information about all the system's GPUs through the
5142 NV-CONTROL X extension. This option can be set to FALSE to disable this
5143 behavior, such that only GPUs with X screens configured on them will be
5144 probed. Default: all GPUs in the system are probed.
5146 Option "DynamicTwinView" "boolean"
5148 Enable or disable support for dynamically configuring TwinView on this X
5149 screen. When DynamicTwinView is enabled (the default), the refresh rate of
5150 a mode (reported through XF86VidMode or XRandR) does not correctly report
5151 the refresh rate, but instead is a unique number such that each MetaMode
5152 has a different value. This is to guarantee that MetaModes can be uniquely
5153 identified by XRandR.
5155 When DynamicTwinView is disabled, the refresh rate reported through XRandR
5156 will be accurate, but NV-CONTROL clients such as nvidia-settings will not
5157 be able to dynamically manipulate the X screen's MetaModes. TwinView can
5158 still be configured from the X config file when DynamicTwinView is
5161 Default: DynamicTwinView is enabled.
5163 Option "IncludeImplicitMetaModes" "boolean"
5165 When the X server starts, a mode pool is created per display device,
5166 containing all the mode timings that the NVIDIA X driver determined to be
5167 valid for the display device. However, the only MetaModes that are made
5168 available to the X server are the ones explicitly requested in the X
5171 It is convenient for fullscreen applications to be able to change between
5172 the modes in the mode pool, even if a given target mode was not explicitly
5173 requested in the X configuration file.
5175 To facilitate this, the NVIDIA X driver will, if only one display device
5176 is in use when the X server starts, implicitly add MetaModes for all modes
5177 in the display device's mode pool. This makes all the modes in the mode
5178 pool available to full screen applications that use the XF86VidMode or
5179 XRandR X extensions.
5181 To prevent this behavior, and only add MetaModes that are explicitly
5182 requested in the X configuration file, set this option to FALSE.
5184 Default: IncludeImplicitMetaModes is enabled.
5186 Option "AllowIndirectPixmaps" "boolean"
5188 Some graphics cards have more video memory than can be mapped at once by
5189 the CPU (generally only 256 MB of video memory can be CPU-mapped). On
5190 graphics cards based on G80 and higher with such a memory configuration,
5191 this option allows the driver to place more pixmaps in video memory which
5192 will improve hardware rendering performance but will slow down software
5193 rendering. On some systems, up to 768 megabytes of virtual address space
5194 will be reserved in the X server for indirect pixmap access. This virtual
5195 memory does not consume any physical resources.
5197 Default: on (indirect pixmaps will be used, when available).
5199 Option "OnDemandVBlankInterrupts" "boolean"
5201 Normally, VBlank interrupts are generated on every vertical refresh of
5202 every display device connected to the GPU(s) installed in a given system.
5203 This experimental option enables on-demand VBlank control, allowing the
5204 driver to enable VBlank interrupt generation only when it is required.
5205 This can help conserve power.
5207 Default: off (on-demand VBlank control is disabled).
5209 Option "PixmapCacheSize" "size"
5211 This option controls how much video memory is reserved for pixmap
5212 allocations. When the option is specified, "size" specifies the number of
5213 bytes to use for the pixmap cache. Reserving this memory improves
5214 performance when pixmaps are created and destroyed rapidly, but prevents
5215 this memory from being used by OpenGL. When this cache is disabled or
5216 space in the cache is exhausted, the driver will still allocate pixmaps in
5217 video memory but pixmap creation and deletion performance will not be
5220 NOTE: This option is deprecated in favor of the PixmapCacheRoundSizeKB
5221 nvidia-settings attribute and will be removed in a future driver release.
5223 Example: "Option "PixmapCacheSize" "1048576"" will allocate one megabyte
5224 for the pixmap cache.
5226 Default: off (no memory is reserved specifically for pixmaps).
5228 Option "AllowSHMPixmaps" "boolean"
5230 This option controls whether applications can use the MIT-SHM X extension
5231 to create pixmaps whose contents are shared between the X server and the
5232 client. These pixmaps prevent the NVIDIA driver from performing a number
5233 of optimizations and degrade performance in many circumstances.
5235 Disabling this option disables only shared memory pixmaps. Applications
5236 can still use the MIT-SHM extension to transfer data to the X server
5237 through shared memory using XShmPutImage.
5239 Default: off (shared memory pixmaps are not allowed).
5241 Option "InitializeWindowBackingPixmaps" "boolean"
5243 This option controls whether the NVIDIA X Driver initializes newly created
5244 redirected windows using the contents of their parent window if the X
5245 server doesn't do it. Leaving redirected windows uninitialized may cause
5246 new windows to flash with black or random colors when some compositing
5247 managers are running.
5249 This option will have no effect on X servers that already initialize
5250 redirected window contents. In most distributions, the X server is patched
5251 to skip that initialization. In this case, it is recommended to leave this
5252 option on for a better user experience.
5254 Default: on (redirected windows are initialized).
5257 ______________________________________________________________________________
5259 Appendix G. Display Device Names
5260 ______________________________________________________________________________
5262 A "display device" refers to some piece of hardware capable of displaying an
5263 image. There are three categories of display devices: analog displays (i.e.,
5264 CRTs), digital displays (i.e., digital flat panels (DFPs)), and televisions.
5265 Note that analog flat panels are considered the same as analog CRTs by the
5266 NVIDIA FreeBSD driver.
5268 A "display device name" is a string description that uniquely identifies a
5269 display device; it follows the format "<type>-<number>", for example: "CRT-0",
5270 "CRT-1", "DFP-0", or "TV-0". Note that the number indicates how the display
5271 device connector is wired on the graphics card, and has nothing to do with how
5272 many of that kind of display device are present. This means, for example, that
5273 you may have a "CRT-1", even if you do not have a "CRT-0". To determine which
5274 display devices are currently connected, you may check your X log file for a
5275 line similar to the following:
5277 (II) NVIDIA(0): Connected display device(s): CRT-0, DFP-0
5279 Display device names can be used in the MetaMode, HorizSync, and VertRefresh X
5280 config options to indicate which display device a setting should be applied
5283 Option "MetaModes" "CRT-0: 1600x1200, DFP-0: 1024x768"
5284 Option "HorizSync" "CRT-0: 50-110; DFP-0: 40-70"
5285 Option "VertRefresh" "CRT-0: 60-120; DFP-0: 60"
5287 Specifying the display device name in these options is not required; if
5288 display device names are not specified, then the driver attempts to infer
5289 which display device a setting applies to. In the case of MetaModes, for
5290 example, the first mode listed is applied to the "first" display device, and
5291 the second mode listed is applied to the "second" display device.
5292 Unfortunately, it is often unclear which display device is the "first" or
5293 "second". That is why specifying the display device name is preferable.
5295 When specifying display device names, you may also omit the number part of the
5296 name, though this is only useful if you only have one of that type of display
5297 device. For example, if you have one CRT and one DFP connected, you may
5298 reference them in the MetaMode string as follows:
5300 Option "MetaModes" "CRT: 1600x1200, DFP: 1024x768"
5303 ______________________________________________________________________________
5305 Appendix H. GLX Support
5306 ______________________________________________________________________________
5308 This release supports GLX 1.4.
5310 Additionally, the following GLX extensions are supported on appropriate GPUs:
5312 o GLX_EXT_visual_info
5314 o GLX_EXT_visual_rating
5320 o GLX_ARB_get_proc_address
5322 o GLX_SGI_video_sync
5324 o GLX_SGI_swap_control
5326 o GLX_ARB_multisample
5328 o GLX_NV_float_buffer
5330 o GLX_ARB_fbconfig_float
5336 o GLX_EXT_texture_from_pixmap
5338 For a description of these extensions, see the OpenGL extension registry at
5339 http://www.opengl.org/registry/
5341 Some of the above extensions exist as part of core GLX 1.4 functionality,
5342 however, they are also exported as extensions for backwards compatibility.
5344 ______________________________________________________________________________
5346 Appendix I. Dots Per Inch
5347 ______________________________________________________________________________
5349 DPI (Dots Per Inch), also known as PPI (Pixels Per Inch), is a property of an
5350 X screen that describes the physical size of pixels. Some X applications, such
5351 as xterm, can use the DPI of an X screen to determine how large (in pixels) to
5352 draw an object in order for that object to be displayed at the desired
5353 physical size on the display device.
5355 The DPI of an X screen is computed by dividing the size of the X screen in
5356 pixels by the size of the X screen in inches:
5358 DPI = SizeInPixels / SizeInInches
5360 Since the X screen stores its physical size in millimeters rather than inches
5361 (1 inch = 25.4 millimeters):
5363 DPI = (SizeInPixels * 25.4) / SizeInMillimeters
5365 The NVIDIA X driver reports the size of the X screen in pixels and in
5366 millimeters. On X.Org 6.9 or newer, when the XRandR extension resizes the X
5367 screen in pixels, the NVIDIA X driver computes a new size in millimeters for
5368 the X screen, to maintain a constant DPI (see the "Physical Size" column of
5369 the `xrandr -q` output as an example). This is done because a changing DPI can
5370 cause interaction problems for some applications. To disable this behavior,
5371 and instead keep the same millimeter size for the X screen (and therefore have
5372 a changing DPI), set the ConstantDPI option to FALSE (see Appendix F for
5375 You can query the DPI of your X screen by running:
5378 % xdpyinfo | grep -B1 dot
5381 which should generate output like this:
5384 dimensions: 1280x1024 pixels (382x302 millimeters)
5385 resolution: 85x86 dots per inch
5389 The NVIDIA X driver performs several steps during X screen initialization to
5390 determine the DPI of each X screen:
5393 o If the display device provides an EDID, and the EDID contains information
5394 about the physical size of the display device, that is used to compute
5395 the DPI, along with the size in pixels of the first mode to be used on
5396 the display device. If multiple display devices are used by this X
5397 screen, then the NVIDIA X screen will choose which display device to use.
5398 You can override this with the "UseEdidDpi" X configuration option: you
5399 can specify a particular display device to use; e.g.:
5401 Option "UseEdidDpi" "DFP-1"
5403 or disable EDID-computed DPI by setting this option to false:
5405 Option "UseEdidDpi" "FALSE"
5407 EDID-based DPI computation is enabled by default when an EDID is
5410 o If the "-dpi" commandline option to the X server is specified, that is
5411 used to set the DPI (see `X -h` for details). This will override the
5412 "UseEdidDpi" option.
5414 o If the "DPI" X configuration option is specified (see Appendix F for
5415 details), that will be used to set the DPI. This will override the
5416 "UseEdidDpi" option.
5418 o If none of the above are available, then the "DisplaySize" X config file
5419 Monitor section information will be used to determine the DPI, if
5420 provided; see the xorg.conf or XF86Config man pages for details.
5422 o If none of the above are available, the DPI defaults to 75x75.
5425 You can find how the NVIDIA X driver determined the DPI by looking in your X
5426 log file. There will be a line that looks something like the following:
5428 (--) NVIDIA(0): DPI set to (101, 101); computed from "UseEdidDpi" X config
5432 Note that the physical size of the X screen, as reported through `xdpyinfo` is
5433 computed based on the DPI and the size of the X screen in pixels.
5435 The DPI of an X screen can be confusing when TwinView is enabled: with
5436 TwinView, multiple display devices (possibly with different DPIs) display
5437 portions of the same X screen, yet DPI can only be advertised from the X
5438 server to the X application with X screen granularity. Solutions for this
5442 o Use separate X screens, rather than TwinView; see Chapter 12 for details.
5444 o Experiment with different DPI settings to find a DPI that is suitable for
5445 both display devices.
5448 ______________________________________________________________________________
5450 Appendix J. XvMC Support
5451 ______________________________________________________________________________
5453 This release includes support for the XVideo Motion Compensation (XvMC)
5454 version 1.0 API on GeForce 6 series and GeForce 7 series add-in cards, as well
5455 as motherboard chipsets with integrated graphics that have PureVideo support
5456 based on these GPUs. There is a static library, "libXvMCNVIDIA.a", and a
5457 dynamic one, "libXvMCNVIDIA_dynamic.so", which is suitable for dlopening.
5458 XvMC's "IDCT" and "motion-compensation" levels of acceleration, AI44 and IA44
5459 subpictures, and 4:2:0 Surfaces up to 2032x2032 are supported.
5461 libXvMCNVIDIA observes the XVMC_DEBUG environment variable and will provide
5462 some debug output to stderr when set to an appropriate integer value. '0'
5463 disables debug output. '1' enables debug output for failure conditions. '2' or
5464 higher enables output of warning messages.
5466 ______________________________________________________________________________
5468 Appendix K. VDPAU Support
5469 ______________________________________________________________________________
5471 This release includes support for the Video Decode and Presentation API for
5472 Unix-like systems (VDPAU) on most GeForce 8 series and newer add-in cards, as
5473 well as motherboard chipsets with integrated graphics that have PureVideo
5474 support based on these GPUs.
5477 K1. IMPLEMENTATION LIMITS
5479 VDPAU is specified as a generic API - the choice of which features to support,
5480 and performance levels of those features, is left up to individual
5481 implementations. The details of NVIDIA's implementation are provided below.
5486 The maximum supported resolution is 4096x4096.
5488 The following surface formats and get-/put-bits combinations are supported:
5490 o VDP_CHROMA_TYPE_420 (Supported get-/put-bits formats are
5491 VDP_YCBCR_FORMAT_NV12, VDP_YCBCR_FORMAT_YV12)
5493 o VDP_CHROMA_TYPE_422 (Supported get-/put-bits formats are
5494 VDP_YCBCR_FORMAT_UYVY, VDP_YCBCR_FORMAT_YUYV)
5500 The maximum supported resolution is 8192x8192.
5502 The following surface formats are supported:
5504 o VDP_RGBA_FORMAT_B8G8R8A8
5506 o VDP_RGBA_FORMAT_R8G8B8A8
5508 o VDP_RGBA_FORMAT_B10G10R10A2
5510 o VDP_RGBA_FORMAT_R10G10B10A2
5512 o VDP_RGBA_FORMAT_A8
5515 Note that VdpBitmapSurfaceCreate's frequently_accessed parameter directly
5516 controls whether the bitmap data will be placed into video RAM (VDP_TRUE) or
5517 system memory (VDP_FALSE). Note that if the bitmap data cannot be placed into
5518 video RAM when requested due to resource constraints, the implementation will
5519 automatically fall back to placing the data into system RAM.
5524 The maximum supported resolution is 8192x8192.
5526 The following surface formats are supported:
5528 o VDP_RGBA_FORMAT_B8G8R8A8
5530 o VDP_RGBA_FORMAT_R10G10B10A2
5533 For all surface formats, the following get-/put-bits indexed formats are
5536 o VDP_INDEXED_FORMAT_A4I4
5538 o VDP_INDEXED_FORMAT_I4A4
5540 o VDP_INDEXED_FORMAT_A8I8
5542 o VDP_INDEXED_FORMAT_I8A8
5545 For all surface formats, the following get-/put-bits YCbCr formats are
5548 o VDP_YCBCR_FORMAT_Y8U8V8A8
5550 o VDP_YCBCR_FORMAT_V8U8Y8A8
5556 In all cases, VdpDecoder objects solely support 8-bit 4:2:0 streams, and only
5557 support writing to VDP_CHROMA_TYPE_420 surfaces.
5559 The exact set of supported VdpDecoderProfile values depends on the hardware
5563 G84, G86, G92, G94, G96, GT200
5565 These chips support the following VdpDecoderProfile values:
5567 o VDP_DECODER_PROFILE_MPEG1, VDP_DECODER_PROFILE_MPEG2_SIMPLE,
5568 VDP_DECODER_PROFILE_MPEG2_MAIN:
5570 o Partial acceleration.
5572 o Minimum width or height: 3 macroblocks (48 pixels).
5574 o Maximum width or height: 128 macroblocks (2048 pixels).
5576 o Maximum macroblocks: 8192
5579 o VDP_DECODER_PROFILE_H264_MAIN, VDP_DECODER_PROFILE_H264_HIGH:
5581 o Complete acceleration.
5583 o Minimum width or height: 3 macroblocks (48 pixels).
5585 o Maximum width or height: 128 macroblocks (2048 pixels).
5587 o Maximum macroblocks: 8192
5592 G98, MCP77, MCP78, MCP79, MCP7A
5594 These chips support the following VdpDecoderProfile values:
5596 o VDP_DECODER_PROFILE_MPEG1, VDP_DECODER_PROFILE_MPEG2_SIMPLE,
5597 VDP_DECODER_PROFILE_MPEG2_MAIN:
5599 o Complete acceleration.
5601 o Minimum width or height: 3 macroblocks (48 pixels).
5603 o Maximum width or height: 128 macroblocks (2048 pixels).
5605 o Maximum macroblocks: 8192
5608 o VDP_DECODER_PROFILE_H264_MAIN, VDP_DECODER_PROFILE_H264_HIGH:
5610 o Complete acceleration.
5612 o Minimum width or height: 3 macroblocks (48 pixels).
5614 o Maximum width: 127 macroblocks (2032 pixels).
5616 o Maximum height: 128 macroblocks (2048 pixels).
5618 o Maximum macroblocks: 8190
5620 o Unsupported widths: 49, 54, 59, 64, 113, 118, 123 macroblocks (784,
5621 864, 944, 1024, 1808, 1888 pixels).
5624 o VDP_DECODER_PROFILE_VC1_SIMPLE, VDP_DECODER_PROFILE_VC1_MAIN,
5625 VDP_DECODER_PROFILE_VC1_ADVANCED:
5627 o Complete acceleration.
5629 o Minimum width or height: 3 macroblocks (48 pixels).
5631 o Maximum width or height: 128 macroblocks (2048 pixels).
5633 o Maximum macroblocks: 8190
5640 The maximum supported resolution is 4096x4096.
5642 The video mixer supports all video and output surface resolutions and formats
5643 that the implementation supports.
5645 The video mixer supports at most 4 auxiliary layers.
5647 The following features are supported:
5649 o VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL
5651 o VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL
5653 o VDP_VIDEO_MIXER_FEATURE_INVERSE_TELECINE
5655 o VDP_VIDEO_MIXER_FEATURE_NOISE_REDUCTION
5657 o VDP_VIDEO_MIXER_FEATURE_SHARPNESS
5659 o VDP_VIDEO_MIXER_FEATURE_LUMA_KEY
5662 In order for either VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL or
5663 VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL to operate correctly, the
5664 application must supply at least 2 past and 1 future fields to each
5665 VdpMixerRender call. If those fields are not provided, the VdpMixer will fall
5666 back to bob deinterlacing.
5668 In order for VDP_VIDEO_MIXER_FEATURE_INVERSE_TELECINE to have any effect, one
5669 of VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL or
5670 VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL must be requested and
5671 enabled. Inverse telecine has the same requirement on the minimum number of
5672 past/future fields that must be provided.
5675 VDPPRESENTATIONQUEUE
5677 The resolution of VdpTime is approximately 10ns. At some arbitrary point
5678 during system startup, the initial value of this clock is synchronized to the
5679 system's real-time clock, as represented by ns since since Jan 1, 1970.
5680 However, no attempt is made to keep the two time-bases synchronized after this
5681 point. Divergence can and will occur.
5683 NVIDIA's VdpPresentationQueue supports two mechanisms for displaying surfaces;
5684 overlay and blit-based. The overlay path will be used wherever possible, with
5685 the blit path acting as a more general fallback. At present, the selection of
5686 overlay v.s. blit path is made at the time of presentation queue creation.
5688 The following conditions or system configurations will prevent usage of the
5691 o Overlay hardware already in use, e.g. by another VDPAU, GL, or X11
5692 application, or by SDI output.
5694 o SLI or Multi-GPU enabled on the given X screen.
5696 o Desktop rotation enabled on the given screen.
5698 o X composite extension enabled on the given screen. Note that simply
5699 having the extension enabled is enough to prevent overlay usage; running
5700 an actual compositing manager is not required.
5702 o The environment variable VDPAU_NVIDIA_NO_OVERLAY is set to a string
5703 representation of a non-zero integer.
5706 At present, the overlay path always syncs to vblank, whereas the blit path
5707 never syncs to vblank.
5710 K2. PERFORMANCE LEVELS
5712 This documentation describes the capabilities of the NVIDA VDPAU
5713 implementation. Hardware performance may vary significantly between cards. No
5714 guarantees are made, nor implied, that any particular combination of system
5715 configuration, GPU configuration, VDPAU feature set, VDPAU API usage,
5716 application, video stream, etc., will be able to decode streams at any
5717 particular frame rate.
5720 K3. GETTING THE BEST PERFORMANCE FROM THE API
5722 System performance (raw throughput, latency, and jitter tolerance) can be
5723 affected by a variety of factors. One of these factors is how the client
5724 application uses VDPAU; i.e. the number of surfaces allocated for buffering,
5725 order of operations, etc.
5727 NVIDIA GPUs typically contain a number of separate hardware modules that are
5728 capable of performing different parts of the video decode, post-processing,
5729 and display operations in parallel. To obtain the best performance, the client
5730 application must attempt to keep all these modules busy with work at all
5733 Consider the decoding process. At a bare minimum, the application must
5734 allocate one video surface for each reference frame that the stream can use (2
5735 for MPEG or VC-1, a variable stream-dependent number for H.264) plus one
5736 surface for the picture currently being decoded. However, if this minimum
5737 number of surfaces is used, performance may be poor. This is because
5738 back-to-back decodes of non-reference frames will need to be written into the
5739 same video surface. This will require that decode of the second frame wait
5740 until decode of the first has completed; a pipeline stall.
5742 Further, if the video surfaces are being read by the video mixer for
5743 post-processing, and eventual display, this will "lock" the surfaces for even
5744 longer, since the video mixer needs to read the data from the surface, which
5745 prevents any subsequent decode operations from writing to the surface. Recall
5746 that when advanced deinterlacing techniques are used, a history of video
5747 surfaces must be provided to the video mixer, thus necessitating that even
5748 more video surfaces be allocated.
5750 For this reason, NVIDIA recommends the following number of video surfaces be
5753 o (num_ref + 3) for progressive content, and no deinterlacing.
5755 o (num_ref + 5) for interlaced content using advanced deinterlacing.
5758 For applications that perform significant amounts of rendering between bitmaps
5759 and output surfaces, a similar argument may apply to the number of output
5762 Finally, consider the display path via the presentation queue. This portion of
5763 the pipeline requires at least 2 output surfaces; one that is being actively
5764 displayed by the presentation queue, and one being rendered to for subsequent
5765 display. As before, using this minimum number of surfaces may not be optimal.
5766 For some video streams, the hardware may only achieve real-time decoding on
5767 average, not for each individual frame. Similarly, system level issues such as
5768 scheduler algorithms and system load may prevent the CPU portion of the driver
5769 from operating for short periods of time. Both of these potential issues may
5770 be solved by allocating more output surfaces, and queuing more than one
5771 outstanding output surface into the presentation queue.
5773 The choice of exactly how many surfaces to allocate is a resource usage v.s.
5774 performance trade-off; Allocating more than the minimum number of surfaces
5775 will increase performance, but use proportionally more video RAM. This may
5776 cause allocations to fail. This could be particularly problematic on systems
5777 with a small amount of video RAM. A stellar application would automatically
5778 adjust to this by initially allocating the bare minimum number of surfaces
5779 (failures being fatal), then attempting to allocate more and more surfaces,
5780 provided the allocations kept succeeding, up to the suggested limits above.
5782 The video decoder's memory usage is also proportional to the maximum number of
5783 reference frames specified at creation time. Requesting a larger number of
5784 reference frames can significantly increase memory usage. Hence it is best for
5785 applications that decode H.264 to request only the actual number of reference
5786 frames specified in the stream, rather than e.g. hard-coding a limit of 16, or
5787 even the maximum number of surfaces allowable by some specific H.264 level at
5788 the stream's resolution.
5790 Note that the NVIDIA implementation correctly implements all required
5791 interlocks between the various pipelined hardware modules. Applications never
5792 need worry about correctness (providing their API usage is legal and
5793 sensible), but simply have to worry about performance.
5796 K4. ADDITIONAL NOTES
5798 Note that surfaces (video, output, or bitmap) are not cleared to any specific
5799 value upon allocation. It is the application's responsibility to initialize
5800 all surfaces prior to using them as input to any function.
5803 K5. DEBUGGING AND TRACING
5805 The VDPAU wrapper library supports tracing VDPAU function calls, and their
5806 parameters. This tracing is controlled by the following environment variables:
5810 Enables tracing. Set to 1 to trace function calls. Set to 2 to trace all
5811 arguments passed to the function.
5815 Filename to write traces to. By default, traces are sent to stderr. This
5816 variable may either contain a plain filename, or a reference to an
5817 existing open file-descriptor in the format "&N" where N is the file
5821 The VDPAU wrapper library is responsible for determining which vendor-specific
5822 driver to load for a given X11 display/screen. At present, it hard-codes
5823 "nvidia" as the driver. The environment variable VDPAU_DRIVER may be set to
5824 override this default. The actual library loaded will be
5825 libvdpau_${VDPAU_DRIVER}.so. Setting VDPAU_DRIVER to "trace" is not advised.
5827 The NVIDIA VDPAU driver can emit some diagnostic information when an error
5828 occurs. To enable this, set the environment variable VDPAU_NVIDIA_DEBUG. A
5829 value of 1 will request a small diagnostic that will enable NVIDIA engineers
5830 to locate the source of the problem. A value of 3 will request that a complete
5831 stack backtrace be printed, which provide NVIDIA engineers with more detailed
5832 information, which may be needed to diagnose some problems.
5834 ______________________________________________________________________________
5836 Appendix L. Tips for New FreeBSD Users
5837 ______________________________________________________________________________
5839 This installation guide assumes that the user has at least a basic
5840 understanding of FreeBSD techniques and terminology. In this section we
5841 provide tips that the new user may find helpful. While the these tips are
5842 meant to clarify and assist users in installing and configuring the NVIDIA
5843 FreeBSD Driver, it is by no means a tutorial on the use or administration of
5844 the FreeBSD operating system. Unlike many desktop operating systems, it is
5845 relatively easy to cause irreparable damage to your FreeBSD system. If you are
5846 unfamiliar with the use of FreeBSD, we strongly recommend that you seek a
5847 tutorial through your distributor before proceeding.
5850 L1. THE COMMAND PROMPT
5852 While newer releases of FreeBSD bring new desktop interfaces to the user, much
5853 of the work in FreeBSD takes place at the command prompt. If you are familiar
5854 with the Windows operating system, the FreeBSD command prompt is analogous to
5855 the Windows command prompt, although the syntax and use varies somewhat. All
5856 of the commands in this section are performed at the command prompt. Some
5857 systems are configured to boot into console mode, in which case the user is
5858 presented with a prompt at login. Other systems are configured to start the X
5859 window system, in which case the user must open a terminal or console window
5860 in order to get a command prompt. This can usually be done by searching the
5861 desktop menus for a terminal or console program. While it is customizable, the
5862 basic prompt usually consists of a short string of information, one of the
5863 characters '#', '$', or '%', and a cursor (possibly flashing) that indicates
5864 where the user's input will be displayed.
5867 L2. NAVIGATING THE DIRECTORY STRUCTURE
5869 FreeBSD has a hierarchical directory structure. From anywhere in the directory
5870 structure, the 'ls' command will list the contents of that directory. The
5871 'file' command will print the type of files in a directory. For example,
5875 will print the type of the file 'filename'. Changing directories is done with
5880 will change the current directory to 'dirname'. From anywhere in the directory
5881 structure, the command 'pwd' will print the name of the current directory.
5882 There are two special directories, '.' and '..', which refer to the current
5883 directory and the next directory up the hierarchy, respectively. For any
5884 commands that require a file name or directory name as an argument, you may
5885 specify the absolute or the relative paths to those elements. An absolute path
5886 begins with the "/" character, referring to the top or root of the directory
5887 structure. A relative path begins with a directory in the current working
5888 directory. The relative path may begin with '.' or '..'. Elements of a path
5889 are separated with the "/" character. As an example, if the current directory
5890 is '/home/jesse' and the user wants to change to the '/usr/local' directory,
5891 he can use either of the following commands to do so:
5897 % cd ../../usr/local
5901 L3. FILE PERMISSIONS AND OWNERSHIP
5903 All files and directories have permissions and ownership associated with them.
5904 This is useful for preventing non-administrative users from accidentally (or
5905 maliciously) corrupting the system. The permissions and ownership for a file
5906 or directory can be determined by passing the -l option to the 'ls' command.
5910 drwxr-xr-x 2 jesse users 4096 Feb 8 09:32 bin
5911 drwxrwxrwx 10 jesse users 4096 Feb 10 12:04 pub
5912 -rw-r--r-- 1 jesse users 45 Feb 4 03:55 testfile
5913 -rwx------ 1 jesse users 93 Feb 5 06:20 myprogram
5914 -rw-rw-rw- 1 jesse users 112 Feb 5 06:20 README
5917 The first character column in the first output field states the file type,
5918 where 'd' is a directory and '-' is a regular file. The next nine columns
5919 specify the permissions (see paragraph below) of the element. The second field
5920 indicates the number of files associated with the element, the third field
5921 indicates the owner, the fourth field indicates the group that the file is
5922 associated with, the fifth field indicates the size of the element in bytes,
5923 the sixth, seventh and eighth fields indicate the time at which the file was
5924 last modified and the ninth field is the name of the element.
5926 As stated, the last nine columns in the first field indicate the permissions
5927 of the element. These columns are grouped into threes, the first grouping
5928 indicating the permissions for the owner of the element ('jesse' in this
5929 case), the second grouping indicating the permissions for the group associated
5930 with the element, and the third grouping indicating the permissions associated
5931 with the rest of the world. The 'r', 'w', and 'x' indicate read, write and
5932 execute permissions, respectively, for each of these associations. For
5933 example, user 'jesse' has read and write permissions for 'testfile', users in
5934 the group 'users' have read permission only, and the rest of the world also
5935 has read permissions only. However, for the file 'myprogram', user 'jesse' has
5936 read, write and execute permissions (suggesting that 'myprogram' is a program
5937 that can be executed), while the group 'users' and the rest of the world have
5938 no permissions (suggesting that the owner doesn't want anyone else to run his
5939 program). The permissions, ownership and group associated with an element can
5940 be changed with the commands 'chmod', 'chown' and 'chgrp', respectively. If a
5941 user with the appropriate permissions wanted to change the user/group
5942 ownership of 'README' from jesse/users to joe/admin, he would do the
5946 # chgrp admin README
5948 The syntax for chmod is slightly more complicated and has several variations.
5949 The most concise way of setting the permissions for a single element uses a
5950 triplet of numbers, one for each of user, group and world. The value for each
5951 number in the triplet corresponds to a combination of read, write and execute
5952 permissions. Execute only is represented as 1, write only is represented as 2,
5953 and read only is represented as 4. Combinations of these permissions are
5954 represented as sums of the individual permissions. Read and execute is
5955 represented as 5, where as read, write and execute is represented as 7. No
5956 permissions is represented as 0. Thus, to give the owner read, write and
5957 execute permissions, the group read and execute permissions and the world no
5958 permissions, a user would do as follows:
5960 % chmod 750 myprogram
5966 The shell provides an interface between the user and the operating system. It
5967 is the job of the shell to interpret the input that the user gives at the
5968 command prompt and call upon the system to do something in response. There are
5969 several different shells available, each with somewhat different syntax and
5970 capabilities. The two most common flavors of shells used on FreeBSD stem from
5971 the Bourne shell ('sh') and the C-shell ('csh') Different users have
5972 preferences and biases towards one shell or the other, and some certainly make
5973 it easier (or at least more intuitive) to do some things than others. You can
5974 determine your current shell by printing the value of the 'SHELL' environment
5975 variable from the command prompt with
5979 You can start a new shell simply by entering the name of the shell from the
5988 and you can run a program from within a specific shell by preceding the name
5989 of the executable with the name of the shell in which it will be run:
5993 The user's default shell at login is determined by whoever set up his account.
5994 While there are many syntactic differences between shells, perhaps the one
5995 that is encountered most frequently is the way in which environment variables
5999 L5. SETTING ENVIRONMENT VARIABLES
6001 Every session has associated with it environment variables, which consist of
6002 name/value pairs and control the way in which the shell and programs run from
6003 the shell behave. An example of an environment variable is the 'PATH'
6004 variable, which tells the shell which directories to search when trying to
6005 locate an executable file that the user has entered at the command line. If
6006 you are certain that a command exists, but the shell complains that it cannot
6007 be found when you try to execute it, there is likely a problem with the 'PATH'
6008 variable. Environment variables are set differently depending on the shell
6009 being used. For the Bourne shell ('sh'), it is done as:
6011 % export MYVARIABLE="avalue"
6013 for the C-shell, it is done as:
6015 % setenv MYVARIABLE "avalue"
6017 In both cases the quotation marks are only necessary if the value contains
6018 spaces. The 'echo' command can be used to examine the value of an environment
6023 Commands to set environment variables can also include references to other
6024 environment variables (prepended with the "$" character), including
6025 themselves. In order to add the path '/usr/local/bin' to the beginning of the
6026 search path, and the current directory '.' to the end of the search path, a
6029 % export PATH=/usr/local/bin:$PATH:.
6031 in the Bourne shell, and
6033 % setenv PATH /usr/local/bin:${PATH}:.
6035 in C-shell. Note the curly braces are required to protect the variable name in
6039 L6. EDITING TEXT FILES
6041 There are several text editors available for the FreeBSD operating system.
6042 Some of these editors require the X window system, while others are designed
6043 to operate in a console or terminal. It is generally a good thing to be
6044 competent with a terminal-based text editor, as there are times when the files
6045 necessary for X to run are the ones that must be edited. Three popular editors
6046 are 'vi', 'pico' and 'emacs', each of which can be started from the command
6047 line, optionally supplying the name of a file to be edited. 'vi' is arguably
6048 the most ubiquitous as well as the least intuitive of the three. 'pico' is
6049 relatively straightforward for a new user, though not as often installed on
6050 systems. If you don't have 'pico', you may have a similar editor called
6051 'nano'. 'emacs' is highly extensible and fairly widely available, but can be
6052 somewhat unwieldy in a non-X environment. The newer versions each come with
6053 online help, and offline help can be found in the manual and info pages for
6054 each (see the section on FreeBSD Manual and Info pages). Many programs use the
6055 'EDITOR' environment variable to determine which text editor to start when
6056 editing is required.
6061 Upon installation, almost all distributions set up the default administrative
6062 user with the username 'root'. There are many things on the system that only
6063 'root' (or a similarly privileged user) can do, one of which is installing the
6064 NVIDIA FreeBSD Driver. WE MUST EMPHASIZE THAT ASSUMING THE IDENTITY OF 'root'
6065 IS INHERENTLY RISKY AND AS 'root' IT IS RELATIVELY EASY TO CORRUPT YOUR SYSTEM
6066 OR OTHERWISE RENDER IT UNUSABLE. There are three ways to become 'root'. You
6067 may log in as 'root' as you would any other user, you may use the switch user
6068 command ('su') at the command prompt, or, on some systems, use the 'sudo'
6069 utility, which allows users to run programs as 'root' while keeping a log of
6070 their actions. This last method is useful in case a user inadvertently causes
6071 damage to the system and cannot remember what he has done (or prefers not to
6072 admit what he has done). It is generally a good practice to remain 'root' only
6073 as long as is necessary to accomplish the task requiring 'root' privileges
6074 (another useful feature of the 'sudo' utility).
6077 L8. FREEBSD MANUAL AND INFO PAGES
6079 System manual or info pages are usually installed during installation. These
6080 pages are typically up-to-date and generally contain a comprehensive listing
6081 of the use of programs and utilities on the system. Also, many programs
6082 include the --help option, which usually prints a list of common options for
6083 that program. To view the manual page for a command, enter
6087 at the command prompt, where commandname refers to the command in which you
6088 are interested. Similarly, entering
6092 will bring up the info page for the command. Depending on the application, one
6093 or the other may be more up-to-date. The interface for the info system is
6094 interactive and navigable. If you are unable to locate the man page for the
6095 command you are interested in, you may need to add additional elements to your
6096 'MANPATH' environment variable. See the section on environment variables.