From LQWiki
Jump to navigation Jump to search

NVIDIA is a company that builds computer hardware and software. Previously most famous because of their GeForce graphic processors and cards, they have since added the nForce mainboard chipsets to their line.

NVIDIA create their own 'nvidia' Linux drivers for most of their hardware but they are closed-source and, for this reason, most distributions install the open source 'nv' drivers after detecting an NVIDIA card. 'nv' drivers are, according to some, easier to install and more stable, but perform a lot worse with 3D because NVIDIA does not disclose the inner workings of their cards.

NVIDIA Graphics Cards installation


To install your NVIDIA graphics card, find out yOur Distribution and proceed accordingly:


Find the "one-click"-procedure for installing your driver at http://en.opensuse.org/Nvidia


For Fedora users, the NVIDIA drivers are available repackaged in RPM, with pre-build kernel modules for easy install and update. Normally you can find the rpms in the livna.org apt/yum respository, but the latest drivers (version 6106) are not yet avaliable as RPMs. See the Fedora FAQ (EL) for more information on how to install on Fedora Core 2 anyway, or here (EL) for more info on the livna nvidia packages.

Drivers before version 6106 will not work with prebuilt Fedora Core 2 kernels, because they do not support 4k Stacks.


Debian has packages for these drivers. There are two packages that need to be installed. One for the binary drivers and libraries and one for the kernel module.
For Debian 3.1 (Sarge) and higher they are called nvidia-glx and nvidia-kernel-<version> (both are in the non-free section). Where <version> is one matching your used Debian kernel or source to build the module for your specific kernel. See the README.Debian file that comes with them for full instructions. And use the package module-assistant to help compile the kernel module if you don't have a default Debian kernel. After installing these packages, the X server configuration file needs to be changed, seen the section about that below.

The other optional packages are nvidia-glx-dev for when you need to compile an openGL applicating against the nvidia openGL libs and nvidia-settings for a GUI application to configure the various driver options.

In Debian 3.0 (Woody) these packages are called nvidia-glx-src and nvidia-kernel-src (both are in the contrib section and use wget to download the actual drivers).

Installing and Configuring nvidia drivers in debian http://www.debianhelp.co.uk/nvidia.htm


Gentoo has ebuilds available which can be installed with:

emerge nvidia-drivers

Once the merge has completed, switch to the nvidia openGL implementation:

eselect opengl set nvidia

You have to edit your X11 config file (/etc/X11/xorg.conf) manually using the procedure described below.

Please use the Official Gentoo documentation on this process before filing a bug (http://www.gentoo.org/doc/en/nvidia-guide.xml)


Mandriva has Nvidia RPMs available for PowerPack users (ie. those who bought a boxed set or are club members). If you have a PowerPack the drivers will install automatically when you install the OS.

For everyone else the manual procedure is:

1) If you haven't already done so go to http://easyurpmi.zarb.org and setup access to the software repositories for your version.

2) In the Software section in the Mandrake Control Centre, or using the urpmi command as root, install the packages "kernel-source-2.6" ("kernel-source" for versions 9.2 or below) and "gcc" (without quotes).

3) Make sure you have the latest kernel - the kernel package isn't updated automatically but the kernel-source package is, which could lead to a version mismatch. To do this run (from a terminal window):

urpmi.update -a
urpmi kernel

If it lists a whole lot of different kernels most users can just select number 1 in the list. Otherwise run

rpm -qa | grep kernel

to see which kernel package you have installed. For more information see http://www.mandriva.com/en/security/kernelupdate

Now follow the rest of the instructions below under "Manual Installation" from "Install" onwards.



The first step is to download some needed files. First, grab the latest version of the drivers from the NVIDIA Website. Once you have that downloaded, make sure you have the source for your kernel. You can check what kernel you have by running the command 'uname -r', without the quotes.


First of all, you need to shut down X. Yes, that's right, we're going to be working from a command line. To do this:

1) Press Ctrl+alt+F1 on the keyboard. Now you should be at a console login screen, login as root.

2) Type the command

telinit 3

to put your machine into runlevel 3. It will tell you about some services its restarting, if it doesn't return you to the command prompt press enter.

Once you get X shut off, cd to where ever you downloaded the driver file to, and run it. You can run it by typing

sh filename

It should check to make sure that it downloaded OK, then start up the installation program.

Go through the first screen, accept the license, and when it asks if you want to try and download a kernel interface from NVIDIA, say no. I've never seen anyone get the kernel interface form the site, so we'll save time and move onto the next screen, where it tells us it will compile the kernel interface on its own. This is where our kernel source comes into play. It should compile and install on its own, so go to the next screen. This should pretty much wrap up anything important in the installation.

Editing the X11 configuration

The next step is to edit your X11 configuration file, so it loads the nvidia driver. For XFree86 the file is usually found in /etc/X11/ and called XF86Config, but you might also have XF86Config-4, in which case you should edit that instead. Many newer distributions use X.org, which uses /etc/X11/Xorg.conf in place of /etc/X11/XF86Config.

Before we do that, we should back up the file, just in case something goes wrong.

# cp /etc/X11/XF86Config /etc/X11/XF86Config.backup

Then, open the X config file in a text editor, e.g.:

# vi /etc/X11/XF86Config

Look for the Module section, and make sure it includes the Load "glx" line. If it doesn't, add it in. Then remove the Load "dri" and Load "GLCore" lines, if you have them. If you might want to go back to the old drivers later you can comment out loading dri and GLCore instead of removing them completely by putting a hash at the start of the line. An example configuration:

Section "Module"
    Load        "dbe"   # Double buffer extension
    Load        "type1"
    Load        "speedo"
    Load        "freetype"
    Load        "glx"
#   Load        "dri"</nowiki> <br>
#   Load        "GLcore"</nowiki> <br>

Finally, find the Device setion, and change the line Driver "nv" or Driver"vesa" to Driver "nvidia". An example device section:

Section "Device"
    Identifier  "NVIDIA"
    Driver      "nvidia"
    VideoRam    65536
    # Insert Clocks lines here if appropriate

To disable the NVidia splash-screen logo when starting X, add the option highlighted below:

Section "Device"
    Identifier  "NVIDIA"
    Driver      "nvidia"
    '''Option      "nologo" "1"'''
    VideoRam    65536
    # Insert Clocks lines here if appropriate

Start X

Now all that's left is to start X back up. To do this type the command 'telinit 5' at the console prompt and it should load X and take you to the normal graphical login screen. You should now have your brand-new NVIDIA drivers installed and working. Good job!

Uninstalling NVIDIA drivers

Did something in your installation go terribly, terrible wrong? No problem! Start your box up in console-only mode, navigate to where you have the nvidia driver installation file saved, and run it with the --uninstall option. sh filename.run --uninstall, and it uninstalls everything it put in. Then it's a simple matter of restoring your XF86Config from the backup we made above.

Card List

GeForce 2 MX 200/400

See also

External links