• Vik@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    6 months ago

    For now, the most effective method to preventing this issue is to limit core clocks on NV31 to about 2500 MHz.

    I believe this is under the known issues for Adrenalin

    • Heavybell@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      edit-2
      6 months ago

      I set mine to 2555 because of Star Citizen (and Ratchet & Clank, but I finished that game a while ago), so I’ve had no issues with HD2 since the first week or so of release. Can confirm it works.

      • Vik@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        6 months ago

        Interesting, were rachet and clank + star citizen also failing in the same way (intermittent TDRs or system shutdown) with default gfx core clocks?

        • Heavybell@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          6 months ago

          In my case it manifested as a graphics driver crash. This would bring down anything that was using the GPU: the game, discord, Windows DWM.

          Someone on SC`s forums did a bunch of research and eventually suggested the clock limit, which has solved all my SC, HD2, R&C issues and most of my War Thunder issues (though that one still silently deletes itself from RAM occasionally, but I think that’s just War Thunder).

          Here’s the post, if you’re interested: https://robertsspaceindustries.com/spectrum/community/SC/forum/50174/thread/high-end-amd-gpu-crashing-possibly-resolved-we-nee

          • Vik@lemmy.world
            link
            fedilink
            English
            arrow-up
            2
            ·
            6 months ago

            Sorry, I take a bunch of acronyms for granted. TDR = timeout detection / recovery.

            Did the bug report tool suggest there was a timeout in each of those cases?

            • Heavybell@lemmy.world
              link
              fedilink
              English
              arrow-up
              2
              ·
              6 months ago

              I forget, it’s been a while. I want to say yes, but the screen would freeze for several seconds before the GPU driver reset in my case, so that seems like a legitimate timeout to me. I had tried increasing the timeout in the past to fix these issues, before I found this method.

              While testing I did confirm using OpenHardwareMonitor that the GPU core frequency would spike above 3100Mhz before crashing, as the thread describes. Setting the max to 2550 seems to prevent this happening.