Announcement

Collapse
No announcement yet.

AMD Stages Latest Radeon/AMDGPU Changes For Linux 4.21 Kernel

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

  • AMD Stages Latest Radeon/AMDGPU Changes For Linux 4.21 Kernel

    Phoronix: AMD Stages Latest Radeon/AMDGPU Changes For Linux 4.21 Kernel

    AMD has posted their initial set of AMDGPU driver changes slated to go into the future Linux 4.21 kernel by way of DRM-Next...

    http://www.phoronix.com/scan.php?pag...nux-4.21-First

  • #2
    EDIT: I'm happy to report that my GPU hangs were related to the stock clocks (when gaming) on the MSI Gaming X RX 480 8GB. After downclocking the sclk from 1303 MHz to 1191 MHz, all is well now and I'm no longer experiencing any hangs. Check this post below for more info.

    ---

    I pray this fixes the GPU hangs with my RX 480. I tried 4.20-rc2 yesterday with the boot parameter amdgpu.gpu_recovery=1 and unfortunately it didn't help when the hang occurred (have to hit the power button). Audio still continues to play though which makes me think it's definitely the GPU hanging. It's happening randomly in _every_ game I try to play. Metro 2033 Redux, Tomb Raider, Rocket League, The Talos Principle, BioShock Infinite, you name it.

    Super frustrating, but still appreciative of all the hard work. I might have to give agd5f's drm-next-4.21-wip branch a whirl today or tomorrow and see if it's fixed.
    Last edited by perpetually high; 11-16-2018, 02:51 AM.

    Comment


    • #3
      Originally posted by perpetually high View Post
      I pray this fixes the GPU hangs with my RX 480. I tried 4.20-rc2 yesterday with the boot parameter amdgpu.gpu_recovery=1 and unfortunately it didn't help when the hang occurred (have to hit the power button).
      Looks like automatic GPU reset is still broken. Did you try the Alt+SysRq / REISUB method at least? It helps me in case of such hangs, and it's better than hardware reset.

      I had to set some flags in Debian though (changing from 438 to 502 in /proc/sys/kernel/sysrq) since it wasn't fully enabled by default. I.e. it was missing "enable signalling of processes" which is E and I in REISUB. For details on flags, see https://www.kernel.org/doc/html/late...ide/sysrq.html
      Last edited by shmerl; 11-14-2018, 05:15 PM.

      Comment


      • #4
        Originally posted by perpetually high View Post
        I pray this fixes the GPU hangs with my RX 480.
        Are you sure it's not a hardware problem, becouse my rx 480 has been very stable for a while?
        Try running it with lower clocks and see if problem goes away.

        Comment


        • #5
          Originally posted by perpetually high View Post
          I pray this fixes the GPU hangs with my RX 480. I tried 4.20-rc2 yesterday with the boot parameter amdgpu.gpu_recovery=1 and unfortunately it didn't help when the hang occurred (have to hit the power button). Audio still continues to play though which makes me think it's definitely the GPU hanging. It's happening randomly in _every_ game I try to play. Metro 2033 Redux, Tomb Raider, Rocket League, The Talos Principle, BioShock Infinite, you name it.

          Super frustrating, but still appreciative of all the hard work. I might have to give agd5f's drm-next-4.21-wip branch a whirl today or tomorrow and see if it's fixed.
          As you I'm in the same situation, it's frustating this happen without reason, the only way I can reproduce it is with heaven benchmark that mades this hang. I will try to compile the kernel with ALT+SysRq, REISUB. But this is not an option, the only I don't get this suddenly happen, just after update of kernel make this situation. Before I played very well with AMDGPU without hangs.

          It's odd to know what make the GPU reset, hanging the system.

          Comment


          • #6
            Originally posted by shmerl View Post

            Looks like automatic GPU reset is still broken. Did you try the Alt+SysRq / REISUB method at least? It helps me in case of such hangs, and it's better than hardware reset.

            I had to set some flags in Debian though (changing from 438 to 502 in /proc/sys/kernel/sysrq) since it wasn't fully enabled by default. I.e. it was missing "enable signalling of processes" which is E and I in REISUB. For details on flags, see https://www.kernel.org/doc/html/late...ide/sysrq.html
            Thanks for the heads up, I didn't know about this. The flags were set to 176 in Ubuntu for me. I've set it to 502 and have been playing around with some the commands and looking at syslog, pretty cool. I'll give this a shot next time the hang happens.

            Originally posted by Etherman View Post
            Are you sure it's not a hardware problem, becouse my rx 480 has been very stable for a while?
            Try running it with lower clocks and see if problem goes away.
            I can't say for certain, but I've set everything in BIOS to stock defaults just to isolate from any CPU overclocking, etc.

            I have the MSI Gaming X RX 480 8GB that has default clocks of 1303/2000 so I leave it at that for gaming with fans on auto. The temperature never gets too high, though.

            Originally posted by raonlinux View Post
            It's odd to know what make the GPU reset, hanging the system.
            I hear you, man. It's starting to give me unnecessary anxiety because I know that GPU hang is coming for me, it's just a matter of when. Seems to happen at a crucial times too, right before an auto-save checkpoint or last few seconds of a Rocket League multiplayer match.

            Comment


            • #7
              Well, is anyone working on fixing kernel 4.19 so AMDGPU works with my R9 390 again? I'm beginning to worry I'll be in the dead zone for another year, again.

              Comment


              • #8
                Took a photo of a GPU hang occurring in Metro 2033 Redux with GALLIUM_HUD env var set:

                https://i.imgur.com/4Dm8RK9.jpg

                - GPU temp: 66c
                - GPU load: 99%
                - CPU's were at 3.6 GHz (Turbo Boost from base 3.4 apparently)
                - CPU loads were 71, 57, 50, 75
                - FPS was at 163
                - VRAM usage was reasonable at 1.175GB

                So the GPU load at 99 is the only thing that sticks out here. Also, you'll also see on the bottom left the textures became screwed up. Usually when that happens, about 1 or 2 seconds later the hang happens, as the case here.

                Comment


                • #9
                  It's nice to see so much work put into AMD drivers, especially the compute unifying part. It would be nice to install one open-source driver and just "magically" get it all. Still waiting for my compute-graphics nirvana...

                  Comment


                  • #10
                    Originally posted by perpetually high View Post
                    Took a photo of a GPU hang occurring in Metro 2033 Redux with GALLIUM_HUD env var set:

                    https://i.imgur.com/4Dm8RK9.jpg

                    - GPU temp: 66c
                    - GPU load: 99%
                    - CPU's were at 3.6 GHz (Turbo Boost from base 3.4 apparently)
                    - CPU loads were 71, 57, 50, 75
                    - FPS was at 163
                    - VRAM usage was reasonable at 1.175GB

                    So the GPU load at 99 is the only thing that sticks out here. Also, you'll also see on the bottom left the textures became screwed up. Usually when that happens, about 1 or 2 seconds later the hang happens, as the case here.
                    I don't game nowadays, but looking into the picture..
                    You have:
                    Requested Vram 948 MB
                    Used Vram 1175 MB

                    Shouldn't it be:
                    Requested VRAM >= 1175 MB?

                    Comment

                    Working...
                    X