Nvidia Videocard Setup In Fedora Linux

[Hlingler]

NOTE: As of 12-15-2008, this How-To will no longer be updated. Please refer to one of the other NVidia How-Tos.

Due to the large number of threads posted recently with questions regarding nvidia video drivers, I thought that perhaps a new/updated FAQ and How-To (or at least a supplement to the existing one) was in order. The original "How-To" by Firewing1, which you may also find useful, as I did, is located here.

THE DIME TOUR...

For those who simply want to install the proprietary NVidia driver as fast as possible with as little effort as possible and (think that they) need nothing more, here's where to go to find the "quickie" install instructions (by the guy who maintains the Livna drivers):

• F10: F10 Nvidia driver guide
• F9: F9 Nvidia driver guide, http://www.firewing1.com/content/view/88/52/
• F8: http://www.firewing1.com/content/view/64/40/

NOTE: Neither the legacy 96xx nor legacy 71xx series drivers support F10/F9's X11/Xorg. So: you won't get 3D/DRI unless you downgrade the whole X11 Window System (F9 only). See the 'How-To' elsewhere in this forum. See A4 below for lists of supported chipsets.

If things don't go as well as you planned and you don't understand why, feel free to wander back here and read a little...

For the rest, or those who want/need a little more info, some FAQS first...

FAQS

Q1: Do I need to install the proprietary nvidia video drivers?
A1: No, you do not need to install the proprietary nvidia video drivers to get a working GUI/desktop (in most cases).

NOTE that Fedora does not supply the proprietary nvidia driver (nor any other proprietary nor patent-encumbered software). You must obtain that driver from some 3rd party if you want/need it (see A5 below).

1. Fedora Linux is supplied with an open-source video driver that works for almost all nvidia chipsets/cards available [anything from RIVA 128 (NV3) to the newest GeForce 8XXX (G80, G84, G86)]. This is the default driver that will be automatically used in the absense of any other driver(s). However, this open-source video driver (named "nv") is a 2D-only driver, so it does not provide 3D hardware-accelerated graphical rendering, nor Direct Rendering (DRI), nor any of the OpenGL protocols needed to do advanced graphical rendering. You will not be able to use some of the available packages (e.g., Compiz-Fusion Compositing Window Manager) with this driver, and your graphics performance may be inadequate for the newer, more demanding games and other applications that require accelerated graphical rendering (e.g., Google Earth). Driver "nv" does provide "accelerated" 2D graphical rendering. Package name: xorg-x11-drv-nv. Driver file/name: /usr/lib/xorg/modules/drivers/nv_drv.so. Documentation: man nv. Web sites: http://dri.freedesktop.org/wiki/NVIDIA, http://www.x.org/wiki/nv
   
   Known Problems: The infamous "invisible mouse cursor" issue; see the Configuration and TroubleShooting sections below for details.
   
   
2. Fedora Linux is also currently supplied with an experimental 2D video driver for nvidia chipsets/cards. It's name is "nouveau". This experimental driver also supports most available nvidia chipsets/cards [anything from RIVA 128 (NV3) to GeForce 7XXX (G70, G71, G72, G73)]. Eventually, this driver is planned to also provide 3D/DRI/OpenGL support (in fact, current kernels ship with the driver kernel module: /lib/modules/<uname -r>/kernel/drivers/char/drm/nouveau.ko, but NOT the actual 3D driver, which would be /usr/lib/dri/nouveau_dri.so), but 3D support is not included at this time. However, for those so inclined, it is possible to acquire the mesa/nouveau source code and compile the 3D driver (Good Luck, you'll need it). As with the "nv" driver, you will not be able to use some of the available packages (e.g., Compiz-Fusion Compositing Window Manager) with this driver, and your graphics performance may be inadequate for the newer, more demanding games and other applications that require accelerated graphical rendering. Driver "nouveau" does provide "accelerated" 2D graphical rendering. Package name: xorg-x11-drv-nouveau. Driver file/name: /usr/lib/xorg/modules/drivers/nouveau_drv.so. Documentation: man nouveau. Web Site: http://nouveau.freedesktop.org/wiki/
   
   Known Problems: Experimental, requires manual configuration to enable; see the Configuration and TroubleShooting sections below for more info.

If you have a sufficiently powerful CPU and/or GPU and/or plenty of VRAM (and system RAM), and you don't want/need/use any graphically-intense applications, and you don't want/need the potential hassle that installing the proprietary drivers can sometimes bring, or you don't want to "taint" your install with proprietary software, then simply use the default open-source drivers that are supplied with Fedora. Chances are that you won't have to configure anything - it will "Just Work™©®". Be aware however, that if you do run any graphically demanding software, your hardware will be working far harder than it needs to, and you will not be getting the best possible performance out of it.

Q2: Is there an open-source video driver for nvidia chipsets/cards that provides 3D/DRI/full OpenGL graphics rendering support?
A2: No. See A1(b).

Q3: I want the best video driver for my nvidia chipset/card so that I can get maximum performance and/or install/use all of the eye-candy and other goodies that I see are available, and I don't care where I get it, as long as it's free. What do I need to do to get that?
A3: Install the proprietary nvidia video driver, appropriate to your flavor of Fedora and your nvidia chipset/card. See the next two FAQs below.

Q4: What nvidia chipsets/cards are supported by the proprietary driver(s)?
A4: There are now FOUR versions of the nvidia driver which support different chipsets/cards:

1. The latest stable release version (currently REV. 177.82) for GeForce 6xxx and newer. List of supported chipsets/cards: http://www.nvidia.com/object/IO_18897.html
2. Series 173.xx for GeForce 5xxx chipsets: http://www.nvnews.net/vbulletin/showthread.php?t=120679
3. The legacy 96xx-series driver. List of supported chipsets/cards: http://www.nvidia.com/object/IO_32667.html
4. The legacy 71xx-series driver. List of supported chipsets/cards: http://www.nvidia.com/object/IO_32667.html or http://www.nvidia.com/object/1.0-718..._products.html

NOTE: Neither the legacy 96xx nor legacy 71xx series drivers support F10/F9's X11/Xorg, so you won't get 3D/DRI unless you downgrade the whole X11 Window System (F9 only). See the 'How-To' elsewhere in this forum. See above links for lists of supported chipsets.

[Hlingler]

FAQS Continued...

Q5: Where do I get the proprietary nvidia video driver(s)?
A5: You can obtain the proprietary nvidia video drivers from several sources (NOTE: there are now FOUR versions which support different chipsets; see A4):

NOTE: Neither the legacy-96xx nor -71xx series drivers support F10/F9's X11, so you won't get 3D/DRI unless you downgrade the whole X11 Window System (F9 only). See the 'How-To' elsewhere in this forum.

NOTE: None of these drivers auto-magically start working the instant after you install them. Depending on where you get the driver, you must, at a minimum, re-start the X-server. Others require that you start a service daemon, and/or re-build the kernel driver module or re-boot the PC. Refer to the instructions accompanying each specific source.

1. From the NVidia Corp. website: http://nvidia.com. Follow the "Download Drivers" link and go through the automated checklist to find the correct version of the driver for your architecture of Fedora (x86 or x86_64) and your nvidia chipset. Be aware that this is a generic executable binary installer that is intended to run on any Linux platform/distribution - it is not optimized for Fedora Linux. Download and install that driver as per the provided instructions/documentation. The driver comes with a GUI configuration utility ('nvidia-settings') to assist you with setup after installation. The Fedora Project provides no technical support for this driver. For help with technical issues, visit the forums here at fedoraforum.org, or nvidia's own user support forums: http://www.nvnews.net/vbulletin/
   
   !!!WARNING!!! Use of the nvidia-provided binary driver installer is strongly discouraged. See: http://fedoraproject.org/wiki/Xorg/3rdPartyVideoDrivers.
   
   NOTE: Also, be aware that this is not a one-time "install-and-forget-about-it" deal - you MUST manually re-install the driver if you update your kernel (NOTE: as an alternative, you can use the custom script provided by forum member brebs to update the kmod for you). You might also need to install the kernel-headers and/or kernel-devel packages in order for the installer to build your driver kernel module (you will be told during the installation procedure if this is required). Also, updates of the driver itself will not be automatic: you will be on your own to check periodically for updated versions of the driver (which fix bugs and/or add new support for new cards/features) and (re)install the driver (or ignore updates at your own risk).
   
   
2. Several reputable 3rd-party software repositories provide re-packaged versions of the same exact nvidia video drivers, in RPM package format, which allow you to install and update the driver with your usual Fedora package management tools (yum, rpm, yumex, kyum, "Package Manager" aka "Pirut", "Software Updater" aka "Pup", "Smart Package Manger", and others). These re-packaged drivers are also custom-built to integrate with maximum compatibility into your Fedora Linux OS. All of the following software repositories provide high-quality re-packaged nvidia video drivers that best suit your Fedora Linux OS. These packages make installing and configuring the nvidia video drivers as easy as possible. Use of the re-packaged nvidia drivers from one of these repos is encouraged.
   
   NOTE: If you have previously installed the nvidia driver using the binary installer from NVidia Corp., you should un-install it before installing any other driver packages.
   
   NOTE: You must select the appropriate package(s) for your version of Fedora and your kernel version (including variations/combinations such as x86, x86_64, PAE, XEN, etc.) as well as the driver suite that supports your chipset (see A4).
   
   !!!WARNING!!! The mixing of software packages from different 3rd-party repos (especially those that provide over-lapping packages, including RPMFusion [Livna+FreshRPMs] and ATRPMs) can frequently lead to package conflicts and update problems (the infamous "repo hell"), and should be avoided, so if you use one set of these packages, you should avoid installing software packages from the other major 3rd-party repos. However, there are various ways to avoid/mitigate these conflicts.
   
   
   
   1. RPMFusion:. A merger of Livna+FreshRPMs+Dribble. Currently the most popular of the 3rd-party repos. In addition to the standard nvidia-settings utility, a special utility named "livna-config-display" is included to assist with setup of the driver. You must also install the kernel-headers, and kernel-devel packages (if you are using yum to install the driver, these other packages should be automatically pulled in also as dependencies). However, the kernel driver module is built at boot time by the akmods utility, so there are no kernel driver modules to update. Don't forget to start the nvidia service daemon, and (for F9 and up) run:

      akmods --kernels `uname -r`

      to re-build the kmod, before re-starting the X-server. Short tutorials to guide you through installing the Livna repo and nvidia driver packages can be found at the following links:
      
      NOTE: Starting with F9, RPMFusion have changed the format of their nvidia package suite to a different model that does not rely on kernel module (kmod) updates for each new kernel. This change should make using this driver suite even easier: there will be no waiting for kmod updates to follow kernel updates. See Post #14 below.
      
      F10: F10 Nvidia driver guide
      F9: F9 Nvidia driver guide, http://www.firewing1.com/content/view/88/52/
      F8: http://forums.fedoraforum.org/showthread.php?t=99285, http://www.firewing1.com/content/view/64/36/, http://www.fedorafaq.org/#nvidia, http://www.mjmwired.net/resources/mj...f8.html#nvidia
      
      Advantages: using the RPMFusion nvidia driver package suite is probably the easiest way to get the proprietary nvidia graphics software working in Fedora Linux. Software updates are automatic with your other usual updates.
      
      Disadvantages: The extra kernel-headers and kernel-devel packages are large and require HHD space for installation; the legacy 71xx-series driver is not available from RPMFusion.
      
      
   2. ATRPMs:. Installation of the driver packages is much the same as with the RPMFusion repo packages, except for the names of the packages in the suite (and there is no special configuration utility). In most ways, this nvidia driver package suite is almost identical to Livna's suite, including the use of kernel modules, but there is no utility to build new kmods - you must update them from ATRPMs. The nvidia-settings utility is of course also included.
      
      Advantages: No other extra nor special packages (e.g., kernel-headers nor kernel-devel) are needed. Software updates are automatic with your other usual updates if you also install package yum-plugin-kmdl from ATRPMs.
      
      Disadvantages: kernel driver module updates may lag several days behind the kernel updates themselves, which frequently leads to GUI problems when impatient users update the kernel and reboot without having updated the driver module (so their graphics interface suddenly "doesn't work anymore"). ATRPMs has in the remote past included configuration files in their repo release RPMs that overwrote Fedora's default files and re-directed yum to use ATRPMs repo. This practice resulted in ATRPMs acquiring a "bad rep" among many users (including yours truly). However, this issue seems to be long in the past, and the packages/files in question are no longer available.

[Hlingler]

FAQS Continued...

Q6: How do I determine if I have Direct Rendering enabled and working?
A6: The easiest way is to open a terminal (as any normal user) and type:

glxinfo|grep render

You want to see:

Direct Rendering: Yes

If you see "No" (or error output), then DRI is either not available with your current driver, or is not working/configured properly for some reason. See "TroubleShooting" below. You should also see another line telling you the name of the chipset/card identified by the software:

[vince@localhost ~]$ glxinfo|grep render
direct rendering: Yes
OpenGL renderer string: RIVA TNT2/PCI/SSE
[vince@localhost ~]$

Q7: How can I test Direct Rendering?
A7: You can also run glxgears as a "quick-and-dirty" guage of your graphics performance. The output of command: glxinfo|grep render will tell you whether or not you have direct rendering enabled and working. However, if you would like to test DRI, open a terminal as any user and command: glxgears; a separate pop-up window will appear with three spinning gears; the output in frames-per-second (FPS) will be shown in the terminal window in 5-second chunks. Run the test with no other load on the system, as it will max out all available resources. Do not hide/cover nor minimize the gearbox. Let it run for 30-60 seconds before stopping it (either close the gearbox window, or CTL-C from the command line). Observe the output.

NOTE: glxgears is not a benchmark tool, and was not designed to accurately measure your system's graphical performance. It is merely a simple a test for direct rendering. If you want a true benchmark of your system's capabilities, install package 'hardinfo' (System Profiler and BenchMark) from the standard Fedora repos and run the FBENCH (FPU RayTracing) test.

Q8: How do I determine what video driver is currently specified in the configuration?
A8: The easiest way is to open a terminal (as any normal user) and type:

cat /etc/X11/xorg.conf|grep -i driver

and observe the output:
Driver "nv" == means that the open-source "nv" driver is being used
Driver "nouveau" == means that the open-source "nouveau" driver is being used
Driver "nvidia" == means that the proprietary "nvidia" driver is being used
If the result is none of these three, you have more serious problems than this How-To can address.

NOTE that this output does not guarantee that the driver is configured and/or working properly....

Q9: Something went wrong, an update borked my setup, the drivers suck and don't work, I can't get <insert name of feature> to work and/or it was working before but now suddenly it isn't, Fedora/Linux SUX! What gives?
A9: In addition to looking at the output of glxinfo and the contents of /etc/X11/xorg.conf (see above), there are several places that you should look for clues as to what went wrong. See the "TroubleShooting" section below.

Q10: How do I use the nvidia-settings utility to adjust/tweak/play with my nvidia driver settings?
A10: Don't, unless you absolutely, positively have no other choice. If you simply cannot resist the urge and must find out what secrets to super-over-clocking your GPU to techno-magically turn it into an eye-popping video-graphical fireball are hidden in that little app (hint: there aren't any), then don't close it with the "Quit" button because that will cause xorg.conf to be re-written, possibly (probably) with changes you didn't want. Instead, close it with the Window Title Bar "X" close button.

• If you found this How-To to be helpful (or not so), please take a moment to rate it.
  
• If you have any questions, comments, suggestions, (constructive) criticisms, or whatever, please feel free to post them here or send me a PM.

Regards,
Hlingler

[Hlingler]

As of Fedora Release 8, most hardware detection and configuration should happen automatically, without user intervention of any kind needed. In many cases however, some manual configuration may be required or desired, including manual editing of the global xorg.conf file. Manual adjustments are more likely if you chose to use the "nouveau" or "nvidia" drivers. The reference manual 'man xorg.conf' has details on the required layout/syntax, and the numerous generic settings available. Some general info/FAQs on Xorg can also be found here: http://www.x.org/wiki/FAQ. Regardless of the driver you use, all configuration of the graphical interface is ultimately done in file /etc/X11/xorg.conf. The GUI configuration utilities described below merely provide convenient ways to edit that file.

NOTE: For dual-head setup, refer also to: http://forums.fedoraforum.org/showthread.php?t=201652 (thanks to member Alecfyz).

NOTE: Starting with F10, the xorg.conf file is not generated by default. If you need to generate one for editing, open a terminal and type:

Xorg -configure :1

which will generate xorg.conf.new in the current directory. As root user, copy it to /etc/X11/xorg.conf (and edit as needed). Then restart X (ALT-CTL-Backspace).

1. "Display Manager" Global changes to the general graphical setup, including (re)generating the xorg.conf file, can be easily done by using the GUI "Display Manager" (aka system-config-display) utility. Root-user privileges/password required. Here, you can setup the Monitor and Videocard type/model, as well as global resolution and color depth, and also dual-head configuration.
   
   NOTE that, unlike other OSs you may have used, Xorg defaults to the highest available resolution and color depth that your hardware can handle - and may disable Direct Rendering or even result in a failed GUI/X-session - unless you manually intervene to change it.
   
   Even if the identified devices do not show your exact make and model of hardware, you should not adjust these settings if everything is working. If things are not working as you like/need and you need to do some changes, then in general if your monitor make and model have not been exactly identified, pick the correct generic type (CRT, LCD, etc.) and the max supported resolution for that monitor, and also the correct make and model of your videocard/chipset. Select default color depth, remembering that "Millions of colors" (aka 24 bpp or "truecolor") is required for much of the eye-candy and other goodies (but "Thousands of colors" aka 16 bpp or "highcolor" will give you twice the graphical performance speed of 24 bpp at the same resolution).
   
   NOTE also that the maximum screen resolution and color depth depend on your available video memory (VRAM), and may limit the settings that you can actually make to something less than the maximum supported by your monitor. Likewise, especially on older and/or on-board chipsets, you may need to reduce the global max screen resolution to obtain the required 24 bpp color depth needed for some compositing effects and other software. Although there is a formula for calculating the VRAM needed for a given screen resolution and color depth, it is as much misleading as informative, because the available VRAM will be used not only to hold the framebuffer (HR * VR * (bpp/8) = Video Memory Used) which is actually used to draw the screen, but also such things as: [AGP]GART aperture/textures, ring/indirect buffers, and other such techno-babble stuff. So far as I can decipher, then, there is no hard-and-fast rule to determine the max possible resolution for a given amount of VRAM. Obviously, the higher the resolution and color depth, the more VRAM will be required. If you have any amount of VRAM less than 64 MB, you will probably have to reduce your global max screen resolution to get 24 bpp color depth. Reducing color depth to 16 bpp will double either the max possible resolution or rendering speed (measured in FPS: frames-per-second), but reduce your display quality and make certain software unable to run.
   
   
2. "livna-config-display" If you installed the nvidia driver suite from Livna/RPMFusion, this utility will be installed as a dependency. Use it at least once (the first time, after you have (re)installed the driver packs) to generate a xorg.conf suitable for your OS. Then, when you are satisfied with your setup, un-check the box that allows livna-config-display to edit xorg.conf. No further auto-edits will be made by livna-config-display unless and until you reverse that choice.
   
   
3. "Desktop Settings" Your personal desktop settings can be further adjusted by opening the appropriate desktop settings utility (varies with desktop). If your desktop's settings utility supports this feature, then reduce display refresh rate to the lowest tolerable setting that does not create "flickering" or "Monitor out-of-sync" errors, to improve performance. Try to keep it above 60Hz. 72Hz is the VESA ergonomic standard. If you're sensitive to those, you need to keep it above that (more info and reference: http://tldp.org/HOWTO/XFree86-Video-...tradeoffs.html).
   
   
4. xrandr Even if your desktop has a display adjustment GUI, it may not support all of the features that you want/need to adjust, or you may simply find it more powerful to use the xrandr utility, one of many handy such tools in the xorg-x11-server-utils package. Some GUI front-ends are also available, like 'krandrtray'. With the xrandr command, you can query and change available screen resolution and refresh rates "on-the-fly":

   [Vince@localhost Sat Jun 07 02:38:22 ~]$ xrandr -q
   Screen 0: minimum 320 x 175, current 1280 x 960, maximum 1280 x 960
   default connected 1280x960+0+0 0mm x 0mm
      1280x960       50.0*
      1152x768       51.0
      1024x768       52.0     68.0     69.0     70.0     71.0
      800x600        53.0     76.0     77.0     78.0     79.0     80.0
      640x480        54.0     87.0     88.0     89.0     90.0
      1280x800       55.0     56.0     57.0
      1280x768       58.0     59.0     60.0
      1280x720       61.0     62.0     63.0
      1152x864       64.0     65.0     66.0     67.0
      960x540        72.0
      840x525        73.0     74.0
      832x624        75.0
      800x512        81.0
      720x450        82.0
      720x400        83.0
      680x384        84.0     85.0
      640x512        86.0
      640x400        91.0     92.0     93.0     94.0
      640x384        95.0     96.0     97.0     98.0
      640x360        99.0    100.0    101.0    102.0
      640x350       103.0
      576x432       104.0    105.0    106.0    107.0
      576x384       108.0
      512x384       109.0    110.0    111.0    112.0    113.0
      416x312       114.0
      400x300       115.0    116.0    117.0    118.0    119.0
      360x200       120.0
      320x240       121.0    122.0    123.0    124.0
      320x200       125.0
      320x175       126.0
   [Vince@localhost Sat Jun 07 02:38:30 ~]$

   Note that the current setting is marked with an asterisk. You can change the screen resolution and refresh rates with, for example:

   [Vince@localhost Sat Jun 07 02:38:30 ~]$ xrandr -s 1024x768 -r 68.0
   [Vince@localhost Sat Jun 07 02:42:43 ~]$

   Changes are applied immediately when you press <Enter>. Read: 'man xrandr' for complete details.

MORE LATER...

[Hlingler]

XORG/DRIVER CONFIGURATION/TWEAKS CONTINUED...

• xorg.conf. There are literally dozens of available settings/adjustments available for Xorg itself, plus driver-specific options. Drivers "nv" and "nouveau" have few specific settings that many users will find helpful. Driver "nvidia" on the other hand has dozens more special settings available, depending on the driver version (latest, legacy-96xx, or legacy-71xx). A generic xorg.conf for the "nv" driver should look much like this:

  # Xorg configuration created by system-config-display
  
  Section "ServerLayout"
  	Identifier     "single head configuration"
  	Screen      0  "Screen0" 0 0
  	InputDevice    "Keyboard0" "CoreKeyboard"
  	InputDevice    "Synaptics" "CorePointer"
  EndSection
  
  Section "InputDevice"
  	Identifier  "Keyboard0"
  	Driver      "kbd"
  	Option	    "XkbModel" "pc105"
  	Option	    "XkbLayout" "us+inet"
  EndSection
  
  Section "InputDevice"
  	Identifier  "Synaptics"
  	Driver      "synaptics"
  	Option	    "Device" "/dev/input/mice"
  	Option	    "Protocol" "auto-dev"
  	Option	    "Emulate3Buttons" "yes"
  	Option	    "SHMConfig" "on"
  EndSection
  
  Section "Device"
  	Identifier  "Videocard0"
  	Driver      "nv"
  EndSection
  
  Section "Screen"
  	Identifier "Screen0"
  	Device     "Videocard0"
  	DefaultDepth     24
  	SubSection "Display"
  		Viewport   0 0
  		Depth     24
  		Modes    "1280x800" "1280x768" "1280x720" "1152x864" "1024x768" "800x600" "640x480"
  	EndSubSection
  EndSection

  NOTE that this is a bare minimum of settings, and was auto-generated using "Display Manager" for one specific machine based on the user's input. In particular, note that in this case, the user chose the max global resolution to be displayed as 1280x800, even though his monitor can support much higher resolutions, due to VRAM limitations. A xorg.conf using the "nouveau" driver would look much the same, except for the name of the driver. Additional options can be added by either manually editing the file (as root user), or using livna-config-display or (not recommended) nvidia-settings, if you installed that driver. A xorg.conf file generated by livna-config-display should look something like this:

  # Xorg configuration created by livna-config-display
  
  Section "ServerLayout"
  	Identifier     "Layout0"
  	Screen      0  "Screen0" 0 0
  	InputDevice    "Keyboard0" "CoreKeyboard"
  	InputDevice    "Mouse0" "CorePointer"
  EndSection
  
  Section "Files"
  	ModulePath   "/usr/lib/xorg/modules/extensions/nvidia"
  	ModulePath   "/usr/lib/xorg/modules"
  	FontPath     "unix/:7100"
  EndSection
  
  Section "Module"
  	Load  "extmod"
  	Load  "glx"
  	Load  "dbe"
  	Load  "extmod"
  EndSection
  
  Section "ServerFlags"
  	Option	    "Xinerama" "0"
  	Option	    "AIGLX" "on"
  EndSection
  
  Section "InputDevice"
  
      # generated from default
  	Identifier  "Mouse0"
  	Driver      "mouse"
  	Option	    "Protocol" "auto"
  	Option	    "Device" "/dev/input/mice"
  	Option	    "Emulate3Buttons" "no"
  	Option	    "ZAxisMapping" "4 5"
  EndSection
  
  Section "InputDevice"
  
      # generated from data in "/etc/sysconfig/keyboard"
  	Identifier  "Keyboard0"
  	Driver      "kbd"
  	Option	    "XkbLayout" "us"
  	Option	    "XkbModel" "pc105"
  EndSection
  
  Section "Monitor"
  
      # HorizSync source: edid, VertRefresh source: edid
   ### Comment all HorizSync and VertSync values to use DDC:
  	Identifier   "Monitor0"
  	VendorName   "Unknown"
  	ModelName    "COMPAQ FS7600"
   ### Comment all HorizSync and VertSync values to use DDC:
  	HorizSync    30.0 - 70.0
  	VertRefresh  50.0 - 160.0
  	Option	    "DPMS"
  EndSection
  
  Section "Device"
  	Identifier  "Videocard0"
  	Driver      "nvidia"
  	VendorName  "NVIDIA Corporation"
  	BoardName   "GeForce 6150 LE"
  	Option	    "AddARGBGLXVisuals" "True"
  EndSection
  
  Section "Screen"
  	Identifier "Screen0"
  	Device     "Videocard0"
  	Monitor    "Monitor0"
  	DefaultDepth     24
  	Option	    "metamodes" "1280x960 +0+0; 1152x768 +0+0; 1024x768 +0+0; 800x600 +0+0; 640x480 +0+0"
  	SubSection "Display"
  		Depth     24
  		Modes    "1600x1200" "1280x1024" "1024x768" "800x600" "640x480"
  	EndSubSection
  EndSection
  
  Section "Extensions"
  	Option	    "Composite" "Enable"
  EndSection

  The most important entries are highlighted in red. NOTE that I have limited my display to 1280x960 using metamodes (due to VRAM and driver considerations, and to improve performance), even though my monitor can support 1600x1200. A xorg.conf file generated by nvidia-settings would look much the same, except the first (commented) lines would read:

  # nvidia-xconfig: X configuration file generated by nvidia-xconfig
  # nvidia-xconfig:  version 1.0  (buildmeister@builder3)  Fri Dec 15 10:12:14 PST 2006

  Again, regardless of actual desktop settings, you must have enough VRAM to support the highest (1st) mode listed, at the default color depth. If you have problems with the xorg.conf auto-generated by either the livna-config-display or nvidia-settings utilities, try these additions to xorg.conf first:

  ModulePath "/usr/lib/xorg/modules/extensions/nvidia"
  ModulePath "/usr/lib/xorg/modules/extensions"
  ModulePath "/usr/lib/xorg/modules"

  to `Section "Files" ' and remove/comment the dri module. Also try adding:

  Section "DRI"
          Mode         0666
  EndSection

• If you found this How-To to be helpful (or not so), please take a moment to rate it.
• If you have any questions, comments, suggestions, (constructive) criticisms, or whatever, please feel free to post them here or send me a PM.

Regards,
Hlingler

MORE LATER...

[Hlingler]

GENERAL TROUBLESHOOTING There are several things that you can do and places to look for clues when something goes wrong:

• Disable unnecessary system services - Graphics-intensive programs can consume enormous amounts of system resources, even with the higher-end hardware nowadays. By first disabling all unnecessary system services/daemons/etc., you can help to ensure that your machine's resources won't be wasted on these other, unseen processes. See mjmuranda's guides: Services in Fedora 9 or Services in Fedora 8. Some system services (e.g., clamav in particular) can suddenly begin to consume huge amounts of system resources, seemimgly without cause, if set up incorrectly (or not set up at all). Also, look for other running/runaway apps/processes that may be consuming system resources. Example: Acrobat Reader has a history of failing to close properly after being started from within FireFox, continuing to run unseen in the background, chewing up almost all CPU power. Use 'top' or a similar process monitor to try to identify such running/runaway apps. The end result is in all cases the same: graphics performance suffers due to lack of system resources.
  
  
• What to do if the entire X11 Window System/GUI is unusable - A lot of incidents occur in which the entire X11/Xorg window system fails to start properly, leaving the user without any functional GUI whatsoever. In such a case, the primary objective will naturally be to restore some kind of GUI. There are two basic cases: (A) No 3rd-party drivers installed; and (B) with 3rd-party drivers installed.
  
  
  
  1. The easiest way to attempt recovery when there are no 3rd-party drivers installed is to get to a Virtual Terminal ("VT") (or (re)boot into run-level 3 = no X11/GUI) and (re)move the /etc/X11/xorg.conf file to a backup location, and then re-start X (or reboot), letting X11/Xorg try to start the GUI with default settings. NOTE that this will not work if you already have problems with the default setup.
     
     
     
     • (Re)boot the machine (or if you feel comfortable doing so, and can do so, login to a VT as root user and simply command:init 3). Stop the GRUB bootloader at the blue screen that appears after the BIOS loads, and follow the on-screen instructions to edit the boot parameters. Add the number '3' to the end of the 'kernel ...' line like so:

       title Fedora (2.6.25.9-40.fc8)
               root (hd0,4)
               kernel /vmlinuz-2.6.25.9-40.fc8 ro root=/dev/sda5 rhgb quiet 3
               initrd /initrd-2.6.25.9-40.fc8.img

       Then proceed with the boot process.
       
       When you get to a VT, log in as root, and command:

       mv /etc/X11/xorg.conf /etc/X11/xorg.conf.screwed-up

       Then reboot, or if you feel comfortable doing so, command:

       init 5

       Hopefully, the machine will start into a functional X11/GUI state.
     
     
     
  2. If you have already installed 3rd-party drivers (from any source), recovery will be a little more difficult because the drivers will always attempt to load unless and until you over-ride them. Follow the above procedure, except that when you get to a VT, you will need to edit xorg.conf to specify a different driver. You will need to use a text-mode editor such as VI (or nano, emacs, or any other text-mode editor). Instead of moving the file, command:

     vi[m] (or: nano, or whatever editor you choose) /etc/X11/xorg.conf

     Inside the text editor, change the line that reads:

     Driver "nvidia"

     to:

     Driver "nv"

     Then re-boot (or command:

     init 5

     Hopefully, the machine will start into a functional X11/GUI state.
  
  
  
• Screen Resolution, Color Depth, and Refresh Rate - As previously stated, X11/Xorg will default to the highest screen resolution, color depth, and refresh rate supported by the monitor(s), even if this results in the disabling of DRI and/or a failed X-session/GUI! Use "Display Manager" to reduce these settings to reasonable values that your VRAM can support.
  
  
• /var/log/Xorg.0.log[.old] - This is the usual system logfile for Xorg X-server. Your syslogd or rsyslog service (or a suitable replacement such as syslog-ng) must be running for any logfiles to be generated. Anything relevant to the running desktop GUI/X-session will be logged to this logfile.
  
  NOTE that this logfile is over-written anew each time a new desktop session is started (whether successfully or not), so any record of previous sessions (including crashes) will be wiped out if you re-start the X-server before inspecting this logfile. Xorg only writes one back-up log (*.old) which is the record of the previous X-session. Ideally, if you are experiencing X-server crashes, try to get to a VT by pressing ALT-CTL-F[1-6] (any one of the six pre-set VTs from F1 through F6 will do) and log in there, then check the logfile and record/backup/save any useful info/data before re-starting the X-server (or trying to...).
  
  As a quick search, you can command:

  grep WW /var/log/Xorg.0.log    <= to sort for warnings
  
  grep EE /var/log/Xorg.0.log    <= to sort for errors

  To save the whole file for later review in your personal home folder:
  

  cp /var/log/Xorg.0.log ~

• ~/.xsession-errors - Your personal xsession logfile might also have some clues (it is also over-written whenever a new desktop session is started):
  

  cat ~/.xsession-errors

  If you are viewing any of the above files in a virtual terminal and/or the text is otherwise scrolling by too fast, then simply append:

  |more

  to the command and use the spacebar to advance to the next screenful of text.
  
  
• Package updates - Software package updates sometimes break things. If you have recently updated packages, or have enabled automatic updates, take a look through /var/log/yum.log also, looking especially for updates to the kernel*, video driver package(s), and/or anything named xorg*. If you don't know what a package is all about, find out:

  rpm -qi <insert-packagename>

  The output will give you a brief summary of the package origin and purpose.

If you have a sufficiently powerful CPU and/or GPU and/or plenty of VRAM (and system RAM), and you don't often use any graphically-intense applications, you may not be aware that a problem has existed for some time before you discovered it. Try not to "jump to conclusions" about what is the root cause of the problem (yeah, I like to do that too ). Rather, take the time to do some background investigation so that, even if you don't find the cause of the problem, then at least you'll have all the info and data when you post your problem to the forum, and you'll have replies to the common responses you'll get.

DRIVER-SPECIFIC ISSUES

1. Driver "nv":
   
   • The infamous "invisible mouse cursor" issue; new installs of Fedora continue to occasionally happen in which the "nv" driver fails to render the mouse cursor - it's there, but not visible.
     Solution: add 'Option "HWCursor" "off" ' and also 'Option "SWCursor" "on" ' to the "Device" section of /etc/X11/xorg.conf (requires root-user privileges).
   
   
   
2. Driver "nouveau":
   
   • F9 KDE fails to start; KDE fails to start from login screen, or aborts back to login screen shortly after opening a desktop session. Bug Report
     Solution: Don't use video driver 'nouveau', use driver 'nv' or some other video driver.
   • F9 KDE fails to start; KDE fails to start from login screen, or aborts back to login screen shortly after opening a desktop session. Bug Report
     Solution: Remove 'Option "BackingStore" "on" (or: "true" or: "1")'
   
   
   
3. Driver "nvidia": Not all of the three available versions of the driver support all options. The older versions support fewer options. Check the documentation that came with the driver software.
   
   SPECIFIC VIDEOCARD/CHIPSET ISSUES
   
   
   1. nVidia 7800GS AGP card with REV. 169.12 driver: Thanx to Jongi for the solution to this issue

      When I try to run games, I get an error: Input signal out of range. I am running an nVidia 7800GS AGP card and this happens with both my F8 and Sidux 64-bit installs. No issues with my Debian 32-bit install. What could be the issue?
      
      Victory with the 169.12 driver. The below needs to be added to the Devices section of the xorg.conf file:

      Option "ModeValidation" "NoXServerModes"

      I was pointed in the right direction when I was trying to solve something Matroska related by switching to a lower resolution. And the reason why the 32-bit Debian worked was because I was running the 100.14 driver there.
      
      This thread provided the solution.

   2. F8: Problem: Driver series 169.* will not work with kernels > 2.6.25*+
      
      Solution: Upgrade to driver series 173.*. The Livna and ATRPMs repos have them. Personally, I downloaded the FreshRPMs F9 SRPM of nvidia-x11-drv-173.14.09-1.beta.fc9 and re-built it against the F8 kernel. The results have been good. If you feel comfortable doing so, you might try that method.

• If you found this How-To to be helpful (or not so), please take a moment to rate it.
• If you have any questions/comments/suggestions/(constructive) criticisms or whatever, please feel free to post here or send me a PM.

MORE LATER...

[Jongi]

Disadvantages: The extra kernel-headers and kernel-devel packages are large and require HHD space for installation; boot time is increased significantly while the kernel driver module is built (time depending of course on the speed of your CPU, etc.), but this only needs to be done once - the kernel module is kept between (re)boots; laptop (and desktop) users who re-start frequently, or those with older, slower CPUs, and those without the patience to wait during the boot process are better served by the driver suites that provide pre-built kernel modules; the legacy 71xx-series driver is not available from FreshRPMs.

I've rebooted about 5 times since installing, and loading the nvidia driver seems to be as quick as when I used kmod-nvidia (which was yesterday). However on first boot after a kernel update, I agree that it does rebuild and that takes longer. Maybe you were saying the same thing. But it came across to me as if you were saying it rebuilds on every reboot.

[leigh123linux]

dkms doesn't take any more time to load here as it seems to build when a new kernel is installed !

[root@localhost ~]# rpm -ivh '/home/leigh/Desktop/kernel-2.6.25.4-23.fc9.x86_64.rpm' '/home/leigh/Desktop/kernel-devel-2.6.25.4-23.fc9.x86_64.rpm' && rpm -Uvh '/home/leigh/Desktop/kernel-headers-2.6.25.4-23.fc9.x86_64.rpm' 
Preparing...                ########################################### [100%]
   1:kernel                 ########################################### [ 50%]
   2:kernel-devel           ########################################### [100%]

nvidia (173.08-1.fc9): Installing module.
...........
........
vboxdrv (1.6.0): Installing module.
..........
......
Preparing...                ########################################### [100%]
   1:kernel-headers         ########################################### [100%]
[root@localhost ~]#

[Firewing1]

Well done, an excellent tutorial! I'm going to try this later today on my laptop and I'll let you know how it goes.
Firewing1

[Xavior]

So how do we install REV. 169.12 ?
I installed the Kernel Dev part, gcc, make, etc... as I usually do to use the nvidia driver from their web,
it's failing, is there a different way to do this now, on F9 ?

thanks,
Xav

[Hlingler]

So how do we install REV. 169.12 ?
I installed the Kernel Dev part, gcc, make, etc... as I usually do to use the nvidia driver from their web,
it's failing, is there a different way to do this now, on F9 ?

thanks,
Xav

See Post #2:

NOTE: NEWS on F9 support: Nvidia driver install for F9
http://www.nvnews.net/vbulletin/showthread.php?t=111460

[BobK1]

I am looking for cookbook-style instructions on how to get the nv driver to work on F9 x86-64bit. I am willing to do without 3D, etc. After a clean install, I can go through the welcome screen, register, then wait for X to load. X hangs, no mouse or keyboard. Have to power down. The F9 distro does work OK.

I have tried adding Disable "glx" in a module section. Have also added Option "IgnoreABI" "true" in a ServerFlags section. I am new to configuring X. I am able to get to single-user mode. Tried startx from there and got "expected keysym, got XF86Info: line 914 of inet". The original hung X log does include these lines:
(WW) "glx" will not be loaded unless you've specified it to be loaded elsewhere.
...
(II) "glx" will be loaded even though the default is to disable it.

I have saved the xorg.conf and log, but not sure how to get them off the machine in single-user mode....

I am running an HP Pavillion d4100e with AMD 64bit dual core 4200 processors, 2GB Ram, 300GB HD. (Same system is dual-boot with XP on a separate HD.)

If this should be posted somewhere else, please move it.
Thanks in advance.

[Hlingler]

@BobK1: Welcome to Fedora. Really.

This is the correct place to post. You can get to a text-mode runlevel 3 by stopping the boot process before it tries to launch the OS:
> Re-start the PC. After BIOS loads, you should see a blank screen very briefly, maybe a few on-screen messages flash by "GRUB loading Stage 2..." (that's the bootloader), then a blue-ish screen appear briefly. The message should say something like "Booting default in X seconds..." You have a very few seconds to press a key to interrupt the next stage. So press any key.
> You can now edit the parameters. Refer to the on-screen instructions: press 'a' to alter the kernel argument; you'll be sent to the end of the second line in the boot orders, where you can change them; add a space then the number '3' (no quotes):

kernel /vmlinuz-2.6.24.7-92.fc8 ro root=/dev/VolGroup00/LogVol00 rhgb quiet 3

Then press enter.
> The OS will boot normally, except it will not attempt to start the X-server/GUI. You will be sent to a full-screen terminal window to log in. Do so, as your normal user.
> At this point, all other services including network should be running - just no GUI. You have several options. Probably your best bet is to edit the /etc/X11/xorg.conf file in a text-mode editor. VI[M] will do.
> Back up the existing xorg.conf file, and the Xorg logfile.. Switch to root user or sudo, and command:

cp /etc/X11/xorg.conf /etc/x11/xorg.conf.screwedup.txt
cp /var/log/Xorg.0.log /var/log/Xorg.0.log.saved.txt

> Switch to root user or start VI with sudo:

[sudo] vi /etc/X11/xorg.conf

VI is a little tricky to get used to, but do the following:
> Use the arrow keys to move around. Go down to the line with "Load glx" and press: dd; the line will be deleted. Do the same with the line with "ignoreABI" and any other custom options etc. that you added. Move to the line where 'Driver "nv" ' is listed, and press the INSERT key; arrow on top of "nv" and press the delete key to delete, and type "vesa" instead.
> If at any point you want to get out of 'edit' mode, press ESC key.
> If at any time you goof and want to quit without changes: ESC, then type :q! and hit enter.
> When you're done and satisfied, to save changes and exit: ESC, then SHIFT-Z-Z.
> Start the X-server as normal user with command: startx

If the X-server loads with drive "vesa", you should get a graphical login screen, where you can proceed as normal. Return here and attach logfile and xorg.conf to new post. Also, what graphics chipset/card is that?

If that fails, return here and post what happened - we'll try something else.

There is no "cookbook" for getting driver "nv" to work - it should automatically configure and work without user action unless something is wrong (case in point).

Logfile should reveal cause(s) of failure.

V

[leigh123linux]

FAQS Continued...



!!!WARNING!!! Use of the nvidia-provided binary driver installer is strongly discouraged. See: http://fedoraproject.org/wiki/Xorg/3rdPartyVideoDrivers.

NOTE: Also, be aware that this is not a one-time "install-and-forget-about-it" deal - you MUST manually re-install the driver if you update your kernel. You might also need to install the kernel-headers and/or kernel-devel packages in order for the installer to build your driver kernel module (you will be told during the installation procedure if this is required). Also, updates of the driver itself will not be automatic: you will be on your own to check periodically for updated versions of the driver (which fix bugs and/or add new support for new cards/features) and (re)install the driver (or ignore updates at your own risk).

NOTE: NEWS on F9 support: 173.08 (beta) for Linux x86/x86-64 released
[Notice] Fedora 9 and Livna, RPMFusion, nVidia+Catalyst drivers
Nvidia driver install for F9

[*]Several reputable 3rd-party software repositories provide re-packaged versions of the same exact nvidia video drivers, in RPM package format, which allow you to install and update the driver with your usual Fedora package management tools (yum, rpm, yumex, kyum, "Package Manager" aka "Pirut", "Software Updater" aka "Pup", "Smart Package Manger", and others). These re-packaged drivers are also custom-built to integrate with maximum compatibility into your Fedora Linux OS. All of the following software repositories provide high-quality re-packaged nvidia video drivers that best suit your Fedora Linux OS. These packages make installing and configuring the nvidia video drivers as easy as possible. Use of the re-packaged nvidia drivers from one of these repos is encouraged.

NOTE: If you have previously installed the nvidia driver using the binary installer from NVidia Corp. web site, you should un-install it before installing any other driver packages.

NOTE: You must select the appropriate package(s) for your version of Fedora and your kernel version (including variations/combinations such as x86, x86_64, PAE, XEN, etc.) as well as the driver suite that supports your videocard/chipset (see A5).

!!!WARNING!!! The mixing of software packages from different 3rd-party repos (especially those that provide over-lapping packages, including Livna, FreshRPMs and ATRPMs) can frequently lead to package conflicts and update problems (the infamous "repo hell"), and should be avoided, so if you use one set of these packages, you should avoid installing software packages from the other major 3rd-party repos. However, there are various ways to avoid/mitigate these conflicts.

1. Livna RPMs: http://rpm.livna.org. Currently the most popular of the 3rd-party repositories. In addition to the standard nvidia-settings utility, a special utility named "livna-config-display" is included to assist with setup of the driver. Short tutorials to guide you through installing the Livna repo and nvidia driver packages can be found at the following links:
   
   NOTE: NEWS on F9 support: http://www.fedoraforum.org/forum/sho...d.php?t=188187
   
   FC4 through F8: http://forums.fedoraforum.org/showthread.php?t=99285
   F8: http://www.firewing1.com/content/view/64/36/, F7: http://www.firewing1.com/content/view/32/36/, FC6: http://www.firewing1.com/content/view/42/36/, FC5: http://www.firewing1.com/content/view/47/36/
   F8: http://www.fedorafaq.org/#nvidia, FC1 through F7: http://www.fedorafaq.org/
   F8: http://www.mjmwired.net/resources/mj...f8.html#nvidia, F7: http://www.mjmwired.net/resources/mj...f7.html#nvidia, FC6: http://www.mjmwired.net/resources/mj...c6.html#nvidia, FC5: http://www.mjmwired.net/resources/mj...c5.html#nvidia, FC4: http://www.mjmwired.net/resources/mj...c4.html#nvidia
   
   Advantages: using the Livna nvidia driver package suite is probably the easiest way to get the proprietary nvidia graphics software working in Fedora Linux. No other extra nor special packages (e.g., kernel-headers nor kernel-devel) are needed. Software updates are automatic with your other usual updates.
   
   Disadvantages: kernel driver module updates may lag several days behind the kernel updates themselves, which frequently leads to GUI problems when impatient users update the kernel and reboot without having updated the driver module (so their graphics interface suddenly "doesn't work anymore").

1. FreshRPMs: http://freshrpms.net/. Installation of the driver packages is much the same as with the Livna repo packages, except that there is only one package (two for x86_64) named nvidia-x11-drv, and you must also install the dkms, kernel-headers, and kernel-devel packages (if you are using yum to install the driver, these other packages should be automatically pulled in also as dependencies). However, the kernel driver module is built at boot time by the dkms utility, so there are no kernel driver modules to update. The standard nvidia-settings utility is of course also included.
   
   Advantages: Since the kernel driver module is built at boot time, there are no kernel driver modules to update. For the impatient or those unwilling to wait for kernel module updates (like me), this re-packaged driver is an excellent choice.
   
   Disadvantages: The extra kernel-headers and kernel-devel packages are large and require HHD space for installation; boot time is increased significantly while the kernel driver module is built (time depending of course on the speed of your CPU, etc.), but this only needs to be done once - the kernel module is kept between (re)boots; laptop (and desktop) users who re-start frequently, or those with older, slower CPUs, and those without the patience to wait during the boot process are better served by the driver suites that provide pre-built kernel modules; the legacy 71xx-series driver is not available from FreshRPMs.
   
   
2. ATRPMs: http://atrpms.net/. Installation of the driver packages is much the same as with the Livna repo packages, except for the names of the packages in the suite (and there is no special configuration utility). In most ways, this nvidia driver package suite is almost identical to the Livna suite, including the use of kernel modules. The nvidia-settings utility is of course also included.
   
   Advantages: No other extra nor special packages (e.g., kernel-headers nor kernel-devel) are needed. Software updates are automatic with your other usual updates if you also install package yum-plugin-kmdl from ATRPMs.
   
   Disadvantages: kernel driver module updates may lag several days behind the kernel updates themselves, which frequently leads to GUI problems when impatient users update the kernel and reboot without having updated the driver module (so their graphics interface suddenly "doesn't work anymore"). ATRPMs has in the remote past included configuration files in their repo release RPMs that overwrote Fedora's default files and re-directed yum to use ATRPMs repo. This practice resulted in ATRPMs acquiring a "bad rep" among many users (including yours truly). However, this issue seems to be long in the past, and the packages/files in question are no longer available.

[/LIST]

Q5: What nvidia chipsets/cards are supported by the proprietary driver(s)?
A5: There are three versions of the nvidia driver which support different chipsets/cards:

1. The latest stable release version (currently REV. 169.12). List of supported chipsets/cards: http://www.nvidia.com/object/IO_18897.html
2. The legacy 96xx-series driver. List of supported chipsets/cards: http://www.nvidia.com/object/IO_32667.html
3. The legacy 71xx-series driver. List of supported chipsets/cards: http://www.nvidia.com/object/IO_32667.html or http://www.nvidia.com/object/1.0-718..._products.html

The livna driver has changed in F9 !
Livna no longer supplies a kmod package compiled against the kernel ( kmod-nvidia is just a meta package now ) , the system has changed to "akmod" ( this is similar to the dkms system ) , the kmod package is built on boot up .

You can also build the kmod package without rebooting with this command

akmods --kernels `uname -r`

So to install the Nvidia driver in F9

su -
rpm -Uhv http://rpm.livna.org/livna-release-9.rpm
yum --enablerepo=livna-testing install  xorg-x11-drv-nvidia xorg-x11-drv-nvidia-libs   akmod-nvidia 
akmods --kernels `uname -r`
nvidia-config-display enable
service nvidia restart

[Hlingler]

OK, thanks for the heads-up, Leigh. I changed the text to account for this new model for F9 and up.

BTW, I've never seen my driver re-build itself during a kernel update like you descibe in Post #8 - is this also perhaps something new for F9 and/or driver REV. 173 ?

V