Announcement

Collapse
No announcement yet.

RadeonSI/Gallium3D Still Appears To Have Greater CPU Overhead Than The NVIDIA Driver

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

  • RadeonSI/Gallium3D Still Appears To Have Greater CPU Overhead Than The NVIDIA Driver

    Phoronix: RadeonSI/Gallium3D Still Appears To Have Greater CPU Overhead Than The NVIDIA Driver

    In CPU-bound Linux games, the NVIDIA Linux driver still appears to perform better than the newest RadeonSI Gallium3D code...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    OMG it's so close!

    Comment


    • #3
      Originally posted by marek View Post
      OMG it's so close!
      Close, but GTX 750 Ti still beats all AMD cards

      I mean newbies who only look at lines probably think that way

      Comment


      • #4
        I wonder if this has been a priority, I might take a look at Mesa since I have some time off, take a profile of it. I think it would be interesting to reduce the CPU usage for other reasons, not just flat-out CPU-bound applications, but also to keep cores running at low frequency for common workloads.

        Comment


        • #5
          Michael I think this test show exactly that also NVIDIA have overhead. Why a 750 perform similar to 750ti and a 1050ti??!!! do you want to measure the over head? You should the test at 640x380 resolution.

          Comment


          • #6
            Originally posted by phoronix View Post
            Phoronix: RadeonSI/Gallium3D Still Appears To Have Greater CPU Overhead Than The NVIDIA Driver

            In CPU-bound Linux games, the NVIDIA Linux driver still appears to perform better than the newest RadeonSI Gallium3D code...

            http://www.phoronix.com/scan.php?pag...1080p-Overhead
            I think it would be interesting to see 99.5th (approx. one per two-hundred frames) percentile and peak frame time. I think the biggest experience improvement can be gained from improvements in peak and high percentile frame time.

            Comment


            • #7
              Originally posted by sp82 View Post
              Michael I think this test show exactly that also NVIDIA have overhead. Why a 750 perform similar to 750ti and a 1050ti??!!! do you want to measure the over head? You should the test at 640x380 resolution.
              960x540 is enough as exactly 1/4 of FullHD and 16:9, some people on Windows with Intel's iGPUs play games on that resolution... it is quite popular

              Michael Why Xonotic isn't also there? That is clear CPU boundware on low too.

              Also i am not sure that ETL is clear CPU bound here with your settings and that "rend2" renderer, with "opengl1" renderer for sure

              It won't be boring to test other paths at least, as both ET:L's rend2 and TF2 tests are GL3 paths CPU bound.
              Last edited by dungeon; 22 May 2017, 09:26 AM.

              Comment


              • #8
                The results are very close considering that there are still some CPU optimization possible for MESA. NVidia driver probably already is very optimized.

                Comment


                • #9
                  The few fps in those benchmarks doesn't matter much and it shows how good the open-source RadeonSI perform, it's not that big of an difference.
                  I'm more exited about the RX 560.

                  Comment


                  • #10
                    I don't think the overhead at 250 fps matters much. You need either a weaker CPU (maybe throttle it) or more demanding benchmarks.

                    Comment

                    Working...
                    X