Announcement

Collapse
No announcement yet.

NVIDIA GeForce RTX 30 Series Linux Driver/Support Expectations

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by Sonadow View Post
    I don't even care if re-clocking works, because it's practically useless for entry-level cards used purely just for displaying the graphical desktop. And I'm still waiting for the day Plasma Wayland can be used with Nouveau. Currently in Debian 10's and OpenSuse Leap 15.1's version of Plasma Wayland, Nouveau:
    - causes an immediately lockup as soon as Plasma Wayland starts
    - locks up after a random amount of time if GL multithreading is disabled
    - locks up after a random amount of time if LIBGL_ALWAYS_INDIRECT is set
    Don't run Wayland on Nvidia. Gnome has an alpha version, KDE is experimental at best, and Sway won't run anyways. None is usable for daily work.
    Better approach: run Wayland desktop on iGPU (or secondary AMD GPU) and use Nvidia for compute&play.

    Comment


    • #22
      I have an nvidia and I regret it.
      It might be a better piece of hardware, but for Linux it sucks. Here are a few reasons:

      0: Closed source in the kernel is always bad. It can have a backdoor that can root my pc for example.
      1: The driver (and the entire stack with the exception of cuda) does not bode well with Linux. It is just ported from windows and it carries burdens from windows.
      2: Code that is not compiled with -march=<target arch> is suboptimal code. Of course nvidia binaries are not compiled with march=zenver1, so they are suboptimal for my pc.
      3: I cannot configure my kernel without kernel module support (and have everything built-in), just because of nvidias modules. This will suck even more when lto is finally mainlined because of the need of exported symbols.

      tl;dr: https://www.youtube.com/watch?v=JbovJbKALzA (I know, we have different reasons to hate nvidia)

      Comment


      • #23
        Originally posted by JackLilhammers View Post
        I want working hardware.
        I know it's not ideal, and I know it goes against my personal beliefs, but at the end of the day I want the GPU to run as best as it can.
        It sucks that Nvidia is so closed, but they still produce the best graphic cards on the market
        We all do but we may see things from a different perspective. For example, I need Nvidia for CUDA/cuDNN. However, having a persistently working and reliable software stack (kernel - driver - application framework, e.g. tensorflow, and application program) with Nvidia proprietary driver is a nontrivial task.

        In practice, you are more or less stuck with a stable distro (Red Hat or Debian stable) and you don't update unless you have to: new kernels can break drivers, drivers break application frameworks and vice versa.
        You end up investing a lot of time if you do update, e.g. because a dev really needs the new tensorflow 2.x functionality or if you decided to try a more up-to-date OS (Fedora or Debian testing) because it "cannot be that bad, right?".

        Then, Nvidia depreciates support after 2-3 generations. After that you are stuck with the last supported kernel, operating system, and software.

        Open source drivers don't fix it all but make your life a lot easier (= breaks less), especially if you run more than 1 box.

        Comment


        • #24
          Originally posted by JackLilhammers View Post

          I want working hardware.
          I know it's not ideal, and I know it goes against my personal beliefs, but at the end of the day I want the GPU to run as best as it can.
          It sucks that Nvidia is so closed, but they still produce the best graphic cards on the market
          if you want working hardware on open source system, vote with your walled and do not purchase for the most hostile and hardware specification hiding vendors. I no-one would have ever purchased Nvidia crap for use with Linux, they would have already been forced to hand us the specification like all normal companies do. https://www.youtube.com/watch?v=el4Byx5dDkw

          Comment


          • #25
            Originally posted by rene View Post

            if you want working hardware on open source system, vote with your walled and do not purchase for the most hostile and hardware specification hiding vendors. I no-one would have ever purchased Nvidia crap for use with Linux, they would have already been forced to hand us the specification like all normal companies do. https://www.youtube.com/watch?v=el4Byx5dDkw
            Nvidia has essentially a monopoly on HPC/compute and, if we are honest, Linux gaming is not a lot of market pressure.
            Both are possibly changing with ROCm/oneAPI and game streaming like Stadia.

            Comment


            • #26
              Originally posted by mppix View Post
              Better approach: run Wayland desktop on iGPU (or secondary AMD GPU) and use Nvidia for compute&play.
              And exactly this is not working. At least I failed to do so on my notebook. With X11 it works surprisingly well.

              Comment


              • #27
                From Ryan Smith original tweet, I was excited that Nvidia would enable SR-IOV for their consumer cards. Turn out that was not the case.

                I actually would buy Nvidia over AMD if SR-IOV was supported in their consumer card (even if they only allow up to one VM guest).

                Comment


                • #28
                  I know people don’t like to admit it but I got a nvidia card and was able to enable wayland on Pop!_OS and gnome actually seems fine! Actually didn’t have some visual glitches my WX2100 was showing. I was surprised cause on kde neon wayland worked so long as you didn’t want a menu or panel...

                  Comment


                  • #29
                    Originally posted by rene View Post
                    Stopp calling binary only blobs valid open source operating system driver support. If you want unfixable binary only stuff use Windows, and stop promoting it for Linux & BSD. As can be seen from the lack of modern API support needed for modern Compositors and desktop environments it is harming innovation and progress.
                    If the driver was open source, the firmware would still be closed. Wuuldn't that be the same kind of problem for the true believers?

                    Comment


                    • #30
                      Originally posted by bemerk View Post
                      If the driver was open source, the firmware would still be closed. Wuuldn't that be the same kind of problem for the true believers?
                      No, it would be a problem only for nutjobs. You know, the people that think hardware with an onboard flash and firmware is OK and a firmware blob loaded by the OS is bad.

                      Comment

                      Working...
                      X