Announcement

Collapse
No announcement yet.

AMDGPU-PRO 16.50 vs. Mesa 13.1-dev + Linux 4.9 Radeon OpenGL

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

  • AMDGPU-PRO 16.50 vs. Mesa 13.1-dev + Linux 4.9 Radeon OpenGL

    Phoronix: AMDGPU-PRO 16.50 vs. Mesa 13.1-dev + Linux 4.9 Radeon OpenGL

    For those curious how the latest open-source AMDGPU+RadeonSI driver code is comparing to yesterday's AMDGPU-PRO 16.50 release, here are some fresh OpenGL Linux driver benchmarks from a few AMD graphics cards.

    http://www.phoronix.com/vr.php?view=23844

  • #2
    Wow, Mesa has really delivered during the last two years when it comes to RadeonSI! Now the only step left is to surpass the AMDGPU Pro driver and reach performance parity with Nvidia

    Comment


    • #3
      Show me a hack to get AMDGPU-Pro to run on Debian Sid with 4.9 Kernel and then I'll be impressed. The rate at which LLVM is limp wristing out 3.9.1 and it's POLARIS fix for broken OpenCL support for Mesa is going on 5 days behind and counting.

      Comment


      • #4
        What is up with Metro Last Light? The Mesa driver is killing AMDGPU-PRO by huge margins. Is the mesa drier rendering correctly?

        Comment


        • #5
          Originally posted by Azpegath View Post
          Wow, Mesa has really delivered during the last two years when it comes to RadeonSI! Now the only step left is to surpass the AMDGPU Pro driver and reach performance parity with Nvidia
          That would be two steps
          An additional step is it needs the goodies in DAL/DC.

          Comment


          • #6
            Originally posted by inhuman4 View Post
            What is up with Metro Last Light? The Mesa driver is killing AMDGPU-PRO by huge margins. Is the mesa drier rendering correctly?
            well that's nothing new, years ago when I had a 7770 the open source driver let me play metro last light (not redux version), with the catalyst on that time even freeze my pc. That was one the reason I just move away from catalyst or closed drivers.

            Comment


            • #7
              I still think the OSS drivers here are highly impressive. Some of the games benched here didn't start out performing this well. I think it is a great validation of AMD's overall approach of make it work first and then make it work fast enough. I don't think AMD needs to match nVidia's driver performance on OpenGL, in fact I don't think they can. The truth is nVidia gets their performance by cheating and I'd rather AMD not stoop to their level. As far as hitting performance characteristics that make sense, AMD is on exactly the right path.

              Comment


              • #8
                Originally posted by duby229 View Post
                in fact I don't think they can. The truth is nVidia gets their performance by cheating and I'd rather AMD not stoop to their level
                what? cheating? [linux source needed]

                Comment


                • #9
                  Originally posted by bug77 View Post
                  That would be two steps
                  An additional step is it needs the goodies in DAL/DC.
                  Huh ? Mesa has nothing to do with DAL/DC other than page flipping via an existing interface.

                  Comment


                  • #10
                    You might be wrong here, bridgman.

                    If you AMD and Kernel Developers get settled about the DAL/DC mess currently going on soon, wouldn't that instantly free up some ressources for tuning other parts of the stack and isn't the current problem going on with DAL/DC a showstopper for other "unrelated" work aligned to it, like bringing up support for new chips?

                    Comment

                    Working...
                    X