Announcement

Collapse
No announcement yet.

A New Set Of OpenGL Benchmarks Come To OpenBenchmarking

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

  • A New Set Of OpenGL Benchmarks Come To OpenBenchmarking

    Phoronix: A New Set Of OpenGL Benchmarks Come To OpenBenchmarking

    FurMark, TessMark, and other advanced OpenGL 2.1/3.2/4.0 benchmarks are now available via the Phoronix Test Suite and OpenBenchmarking.org...

    http://www.phoronix.com/vr.php?view=MTM3MDM

  • #2
    32 bit?

    Did I oversee something or is there no 32 bit version of those benchmarks?

    Comment


    • #3
      Great feature!! But I have more serious trouble with PTS... I don't understand how can I find list of uploaded results based on my criteria efficiently.

      For example,.. how to find all results for driver "radeon" with profile "Unigine Sanctuary" sorted newest first ? :/

      So for me, PTS is remarkably flexible testing solution, but its remarkably hard to compare them to existing, because finding similar is hard ...

      Comment


      • #4
        FYI, FurMark, Pixmark Piano, and Pixmark Volplosion all run on i965 with Mesa master if you do:
        export MESA_GL_VERSION_OVERRIDE=3.2
        export MESA_GLSL_VERSION_OVERRIDE=150

        So they ought to work properly with the next major release of Mesa.

        gimark and tessmark still don't work. I presume this is because they rely on geometry or tesselation shaders, but I haven't looked into the details.
        Free Software Developer .:. Mesa and Xorg
        Opinions expressed in these forum posts are my own.

        Comment


        • #5
          Nice this is actually pretty exciting! Furmark is my go-to program for testing GPU overclocks and it has always annoyed me that I had to install Windows for 1 test.



          Michael:
          You should create a live CD for your PTS, based on some rolling release distro like gentoo or arch (so it's less to maintain). For many people like myself, installing the entire suite isn't really necessary and could actually be kinda a pain if we wanted to test it on multiple platforms. Maybe it could come with a few extra diagnostics like stress tests, sensor monitors, SMART tools, etc.

          Comment


          • #6
            Originally posted by schmidtbag View Post
            Nice this is actually pretty exciting! Furmark is my go-to program for testing GPU overclocks and it has always annoyed me that I had to install Windows for 1 test.



            Michael:
            You should create a live CD for your PTS, based on some rolling release distro like gentoo or arch (so it's less to maintain). For many people like myself, installing the entire suite isn't really necessary and could actually be kinda a pain if we wanted to test it on multiple platforms. Maybe it could come with a few extra diagnostics like stress tests, sensor monitors, SMART tools, etc.
            Bad idea. The CD or USB port would become the bottleneck, and if you copied the entire distro to RAM then it would affect any RAM-sensative benchmarks

            Comment


            • #7
              Originally posted by Kayden View Post
              FYI, FurMark, Pixmark Piano, and Pixmark Volplosion all run on i965 with Mesa master if you do:
              export MESA_GL_VERSION_OVERRIDE=3.2
              export MESA_GLSL_VERSION_OVERRIDE=150
              unfortunately same can not be said about r600 driver just yet. Setting those options for radeon driver made my cinnamon unbootable with Mesa master

              Comment


              • #8
                Doesn't seems to work for me:
                phoronix-test-suite benchmark 1305113-PTS-GPUTESTI20

                ERROR: You must enter at least one test, suite, or result identifier to run.

                Comment


                • #9
                  Originally posted by Ericg View Post
                  Bad idea. The CD or USB port would become the bottleneck, and if you copied the entire distro to RAM then it would affect any RAM-sensative benchmarks
                  I don't see how the CD or USB port would be a bottleneck unless the benchmark needs to reference the disk every once in a while in the middle of it's tests, in which case that particular benchmark could be cached before being run. It's a better idea than loading the entire image.

                  Comment


                  • #10
                    Originally posted by schmidtbag View Post
                    Michael:
                    You should create a live CD for your PTS, based on some rolling release distro like gentoo or arch (so it's less to maintain). For many people like myself, installing the entire suite isn't really necessary and could actually be kinda a pain if we wanted to test it on multiple platforms. Maybe it could come with a few extra diagnostics like stress tests, sensor monitors, SMART tools, etc.
                    I've already done such: http://www.phoronix.com/scan.php?pag...S+Desktop+Live

                    I do have a much more compelling version to be publicly released before EOY assuming I find some more server mirrors. Since the last public PTS Desktop Live release I've done some custom engineering projects on it with more features but lack of bandwidth / mirrors has kept me from releasing them.

                    There's also similar projects using PTS in this way like Gears On Gallium: http://www.phoronix.com/scan.php?pag...tem&px=MTM1ODA
                    Michael Larabel
                    http://www.michaellarabel.com/

                    Comment


                    • #11
                      Testing on Linux and Windows

                      I've tested gputest on my laptop on Kubuntu 12.10 and Windows 7.

                      Everything was tested fullscreen (1920x1080)
                      On Kubuntu, I used the phoronix test suite and on windows the gputest gui.

                      Both Kubuntu and Windows used the 13.4 amd drivers.

                      Gimark:
                      intel hd 4000 (windows): 382
                      intel hd 4000 (kubuntu): didn't work
                      amd hd 7730m (windows): 1700
                      amd hd 7730m (kubuntu): 1728

                      Furmark:
                      intel hd 4000 (windows):240
                      intel hd 4000 (kubuntu): didn't work
                      amd hd 7730m (windows):586
                      amd hd 7730m (kubuntu): 568

                      Tessmark:
                      intel hd 4000 (windows): didn't work
                      intel hd 4000 (kubuntu): didn't work
                      amd hd 7730m (windows): didn't work
                      amd hd 7730m (kubuntu): 2150

                      Piano:
                      intel hd 4000 (windows): 64
                      intel hd 4000 (kubuntu): didn't work
                      amd hd 7730m (windows): 90
                      amd hd 7730m (kubuntu): 94

                      Volplosion:
                      intel hd 4000 (windows): 138
                      intel hd 4000 (kubuntu): didn't work
                      amd hd 7730m (windows): 246
                      amd hd 7730m (kubuntu): 255


                      It confirms that amd drivers performance is similar between windows and linux.
                      It shows also that the ratio (directx performance for hd4000)/(directx performance for hd7730m) is the same than (opengl 3 performance for hd4000)/(opengl 3 performance for hd7730m)

                      That stops there, because aside this benchmark, opengl 1 is buggy for me on windows with the amd card and perfect with the intel card
                      and on kubuntu, portal has lags with the amd card and the games I play with wine work far better with the intel card.

                      Comment

                      Working...
                      X