No announcement yet.

last night i try a nvidia-driver on 2.6.33 and fail

  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    It is possible to install the nvidia drivers while running X. (And come on, this is not Windows - anything can be done any time .)

    If you are using the installer from the website, running the following would have given you what you wanted:
    sh --help
    sh --advanced-options
    sh --no-x-check
    . My other favorite options are "-K", i.e., "--kernel-module-only" and "-k", i.e., "--kernel-version".
    Make sure to add the line
    Driver "nvidia
    in Device section of your xorg.conf and then,
    /etc/init.d/gdm restart
    should restart X with green drivers.

    Btw, I am on Gentoo and I just use the distro package ("emerge nvidia-drivers") instead of the installer. Mainly because it has a rather awesome and sophisticated way of linking the opengl libraries. If Debian and/or Fedora or any distro ever get the equivalent of "eselect opengl", let me know .


    • #12
      This is only partly correct. Certainly you can install driver while X is running, that can be done using

      DISPLAY= nohup sh

      too with my script. At least my script tries to unload the nvidia kernel module (from an older driver) first and also it enables dkms. I certainly know that you can use -aKk kernel-ver but dkms is much smarter... Without unloading old nvidia modules first you have to reboot.


      • #13
        Doesn't nvidia have some kind of package build system like fglrx?


        • #14
          Nope, i tried with a debian stub to do that, but it is no real fun because every few driver releases you needed to adjust it. That's not that problematic when you do that for a single driver manually but when you want to do that automatically for future releases you will not have so much fun. I stopped the idea after a few tries. EnvyNG did just that what you wanted, no idea if it still works - like 195.xx series has additional libs not there in older versions, but maybe it is more tricky than my approach - i did not use it


          • #15
            He seems to be trying to make this hard.
            1. Download the driver.
            2. Make it executable.
              (If you're using Nautilus you could right click the downloaded file and change permissions that way, or use the chmod command.)
            I've covered the above proceedure, but for Kubuntu here. (Look for Steps 2 & 3.) You should be able to generalize these instructions across to Gnome.

            Having done the above, the next stage is to actually run the installer. Remember, if you didn't download the driver into the home directory, you should cd to the directory where the driver file is before running it.
            1. Press CTRL-ALT-F1 (This takes you to a text console.)
            2. Login with an administrator account. The one created with the Ubuntu installer will do.
            3. type sudo stop gdm (You'll need to re-enter your password again after this)
            4. type sudo ./
              Then you need to go through the installer, following the prompts.
              (Remember, you should let the installer modify xorg.conf when it asks if this system has never had the nVidia blob installed.)

            If all goes well, it will tell you so, and if it doesn't you can upload the installer log file here so that we can analyze why it failed, and formulate a solution.

            Assuming that the installer succeeded, you would type:
            sudo start gdm

            and hey, presto you should have a graphical log-in screen. Log in and enjoy the new driver.


            • #16
              Why 195.36.03 when 195.36.08 is the latest stable? The manual way you never get dkms support - until you know how do do that, i do


              • #17
                Ubuntu also has an nvidia ppa. Just google for it. Add that as a software source source and suddenly your video driver is managed by apt, no scripts or advanced juju needed.

                Installs the latest 195 driver which worked flawlessly for me on 9.10. Combine that with the xorg crack pushers ppa and use the latest mainline kernel for that true on the edge experience + userland package sanity AMD users can only dream of.


                • #18
                  That's also an idea, just that i can update my script in about 1 min after i know of a new driver - i just prefer to test that driver before.


                  • #19
                    Originally posted by Kano View Post
                    That's also an idea, just that i can update my script in about 1 min after i know of a new driver - i just prefer to test that driver before.
                    Oh, I'm not saying your script isn't a good idea -- it'll work on anything. Arch, Fedora, oddball foreign releases... But for your average Ubuntu desktop user the ppa may be simpler and more trouble free.

                    My point was to underscore another diffrence between NV and ATI when it comes to Linux drivers. NV proactively manages their own ubuntu PPA to provide their users with an easy, trouble free access to the latest drivers (if that's what they desire) whereas with ATI you're at the mercy of your distribution's vendor. Which is to say, your only hope is community support.

                    I fully expect to be enjoying Xorg 1.8 with fresh nv drivers, KMS and a modern kernel on my desktop without ever having to fire up a shell prompt on my NV powered boxes.


                    • #20
                      When you use -m option and you maybe optimize the code how to shutdown/start X which is different from distro to distro it should work when build depends for the kernel module is installed, thats correct. Also the dkms code could be written a bit more generic, i currently create a deb and install that, thats not really needed, but more easy to uninstall a nvidia-dkms package than using a dkms command - which is not well known for most ppl. You could create rpm too with dkms btw.