• OwOarchist@pawb.social
    link
    fedilink
    English
    arrow-up
    27
    arrow-down
    2
    ·
    6 hours ago

    When you want to do GPU processing for AI, crypto, video editing, etc, though, this gets reversed.

    Getting Cuda working on Linux with an nvidia card is relatively painless. Just a few well-documented commands, worked on the first try.

    I could never get AMD’s equivalent to work on Linux, though, and it led me down a horrible rabbit-hole of trying a dozen different driver versions from a dozen different places, all with their own unique and quirky ways of installing… And it still never did work.

    • brucethemoose@lemmy.world
      link
      fedilink
      arrow-up
      2
      ·
      edit-2
      2 hours ago

      Thats just poor distro support, kind of like CUDA in the past… ROCM should “just work” if it’s shipped right. But it’s not really a priority with maintainers.

      Now, if you’re trying to run CUDA stuff with ROCM, that’s a whole different story. The bast majority of GPU software has extremely poor ROCM support compared to CUDA, and some of this is definitely from AMD footgunning.

    • juipeltje@lemmy.world
      link
      fedilink
      arrow-up
      5
      arrow-down
      1
      ·
      5 hours ago

      For me it was deadsimple once i tried setting it up with nix, granted you need to learn a little about nix so maybe that cancels it out a bit lol.

    • Kazumara@discuss.tchncs.de
      link
      fedilink
      arrow-up
      6
      ·
      5 hours ago

      For me cuda was painful. I did the well documented commands, rebooted and had no output on my laptop screen anymore. Probably a complication due to Optimus, but still…