Announcement

Collapse
No announcement yet.

AMDGPU vs. Radeon DRM With GCN 1.0/1.1 GPUs On Linux 4.11, Mesa 17.1-dev

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

  • #11
    Originally posted by smitty3268 View Post

    Just look at the article, 2 of the 4 GPUs tested show performance problems and 2 of them had problems with output as well requiring Michael to treat them carefully. Plus 1 test caused crashes.

    I'd say that's 3 reasons right there to keep it marked experimental.
    Plus the previously stated lack of UVD support and also hdmi audio support.
    I hope there's nothing more

    Comment


    • #12
      Originally posted by smitty3268 View Post

      Just look at the article, 2 of the 4 GPUs tested show performance problems and 2 of them had problems with output as well requiring Michael to treat them carefully. Plus 1 test caused crashes.

      I'd say that's 3 reasons right there to keep it marked experimental.
      it looks gcn 1.1 sea islands is ready for AMDGPU drm. I.e. R9 290, 290x, 390, 390x etc. Hope it's enabled by default in kernel 4.13
      Last edited by humbug; 07 April 2017, 01:08 AM.

      Comment


      • #13
        Originally posted by atomsymbol
        UVD is supported by amdgpu.ko on R9 290/390[…][/url]
        That's great, so for the GCN 1.1 amdgpu is more or less the preferred choice
        Last edited by Namenlos; 09 April 2017, 01:22 PM. Reason: removed false information about hdmi audio

        Comment


        • #14
          Does anyone know if the bug (below) with Hitman (2016) on a R9 280 is also triggered using amdgpu?

          Comment


          • #15
            Originally posted by atomsymbol

            I don't understand why people think that HDMI audio does not work with GCN 1.1 and amdgpu.ko. Pulseaudio HDMI output works fine with my R9 390.

            Can somebody with a GCN 1.1 AMD GPU installed in the machine confirm that HDMI audio does not work? Maybe some GCN 1.1 GPUs are supported and others not.
            I have the A10-7860K, which is an APU with GCN 1.1 and I do not confirm that HDMI audio does not work, since it works.

            Comment


            • #16
              Originally posted by atomsymbol

              I don't understand why people think that HDMI audio does not work with GCN 1.1 and amdgpu.ko. Pulseaudio HDMI output works fine with my R9 390.

              Can somebody with a GCN 1.1 AMD GPU installed in the machine confirm that HDMI audio does not work? Maybe some GCN 1.1 GPUs are supported and others not.
              So if my memory is correct, a few months ago Alex said that some audio code, for CK I think but not exactly sure anymore, had been ported to amdgpu but not really tested much and so some cards might get it working. Maybe it works for all 1.1? Definitely not for all 1 at least.

              Comment


              • #17
                Originally posted by humbug View Post
                it looks gcn 1.1 sea islands is ready for AMDGPU drm. I.e. R9 290, 290x, 390, 390x etc. Hope it's enabled by default in kernel 4.13
                Do you say it's ready based on some news (as opposed to this benchmark)? Has anything evolved in this scenario since the test was carried on?

                (edit) Are VCE and UVD supported now?
                Last edited by Marc.2377; 20 May 2017, 07:46 PM.

                Comment


                • #18
                  Originally posted by Marc.2377 View Post
                  (edit) Are VCE and UVD supported now?
                  VCE and UVD have always been supported for CI. They are only missing for SI.

                  Comment


                  • #19
                    Originally posted by agd5f View Post

                    VCE and UVD have always been supported for CI. They are only missing for SI.
                    Yeah, I noticed it's marked "not implemented" in the source. Naive question, but do you think it's feasible for someone to cleanly implement them for SI based on existing CIK(amdgpu) and SI(radeon) code?

                    Comment


                    • #20
                      Originally posted by Marc.2377 View Post

                      Yeah, I noticed it's marked "not implemented" in the source. Naive question, but do you think it's feasible for someone to cleanly implement them for SI based on existing CIK(amdgpu) and SI(radeon) code?
                      Yes I do.

                      Comment

                      Working...
                      X