Been using AMD gpus for years due to their superb Linux support. Would like to play around with AI image generation completely offline so a Nvidia gpu is kinda must. Besides that I do some gaming as well.

Currently I have AMD RX 6750 XT and am considering replacing it with a RTX 5060 TI with 16 GB vram. How well does Nvidia gpus work in modern Linux systems. I’m running Tumbleweed with KDE plasma on X11.

Other hardware is Ryzen 7 7700X and an Asrock B650 PG Lightning motherboard.

Edit: Been using InvokeAI so far but cpu only is super slow.

    • FrostyPolicy@suppo.fiOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      3 months ago

      Wish it was that simple. If I select AMD device to use it won’t work. It crashes with: RuntimeError: HIP error: invalid device function.

      Edit: Have a feeling it might be trying to use the igpu for some reason.

        • FrostyPolicy@suppo.fiOP
          link
          fedilink
          English
          arrow-up
          1
          ·
          3 months ago

          Did u run

          rocm-smi
          

          It gives:

          ========================================== ROCm System Management Interface ==========================================
          ==================================================== Concise Info ====================================================
          Device  Node  IDs              Temp    Power    Partitions          SCLK    MCLK     Fan  Perf  PwrCap  VRAM%  GPU%
          (DID,     GUID)  (Edge)  (Avg)    (Mem, Compute, ID)
          ======================================================================================================================
          0       1     0x73df,   60335  40.0°C  8.0W     N/A, N/A, 0         500Mhz  96Mhz    0%   auto  220.0W  19%    0%
          1       2     0x164e,   6110   39.0°C  29.047W  N/A, N/A, 0         N/A     3000Mhz  0%   auto  N/A     4%     0%
          ======================================================================================================================
          ================================================ End of ROCm SMI Log =================================================
          
        • FrostyPolicy@suppo.fiOP
          link
          fedilink
          English
          arrow-up
          1
          ·
          3 months ago

          You’re a life saver! Did the force reinstall mentioned in that github issue and ran Invoke with the command you provided. Now it runs with gpu rendering 🥳. Though I ran out of vram 🤣.