Announcement

Collapse
No announcement yet.

AMD Radeon Gallium3D More Competitive With Catalyst On Linux

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

  • AMD Radeon Gallium3D More Competitive With Catalyst On Linux

    Phoronix: AMD Radeon Gallium3D More Competitive With Catalyst On Linux

    With the ever-changing state of Linux graphics drivers -- both for the open and closed-source drivers -- new tests have been conducted to compare the OpenGL graphics performance on Linux with AMD Radeon graphics. In this article are benchmarks of nine different Radeon HD graphics cards when being tested on the very latest AMD Catalyst (13.3 Beta 3) graphics driver as well as the open-source AMD Radeon driver consisting of Mesa 9.2-devel and the yet-to-be-released Linux 3.9 kernel.

    http://www.phoronix.com/vr.php?view=18654

  • #2
    Interesting that the 6450 in particular has some anomalies. Anyone possible reason why?

    Comment


    • #3
      Michael, if you tested Vadim's shader optimizer http://cgit.freedesktop.org/~vadimg/mesa/log/?h=r600-sb,
      results could have been even better!

      Comment


      • #4
        Maybe I missed it... were the cards running dynpm, medium (default) or high frequencies?

        Comment


        • #5
          It would be really great if Michael could maintain a dynamic feature table comparing all the available linux open and closed graphic drivers. Openbenchmarking does this for performance, but, as development is speeding up, it is really hard to know exactly which hardware features are supported on which driver.

          Comment


          • #6
            Originally posted by Figueiredo View Post
            It would be really great if Michael could maintain a dynamic feature table comparing all the available linux open and closed graphic drivers. Openbenchmarking does this for performance, but, as development is speeding up, it is really hard to know exactly which hardware features are supported on which driver.
            But that's what the freedesktop wiki is for.

            Comment


            • #7
              Originally posted by Ericg View Post
              Maybe I missed it... were the cards running dynpm, medium (default) or high frequencies?
              The default isn't Medium, it's whatever is set in the video cards BIOS.

              Comment


              • #8
                Originally posted by Figueiredo View Post
                It would be really great if Michael could maintain a dynamic feature table comparing all the available linux open and closed graphic drivers. Openbenchmarking does this for performance, but, as development is speeding up, it is really hard to know exactly which hardware features are supported on which driver.
                I think floating point textures/GL3 was disabled, since the specs in the article only mentioned GL2. But S2TC should have been available, since i think Ubuntu ships that by default.

                Comment


                • #9
                  Originally posted by Ericg View Post
                  Maybe I missed it... were the cards running dynpm, medium (default) or high frequencies?
                  The used frequencies are shown automatically by PTS (when debugfs is readable) in the system table, which are shown on the second page.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #10
                    Originally posted by smitty3268 View Post
                    I think floating point textures/GL3 was disabled, since the specs in the article only mentioned GL2. But S2TC should have been available, since i think Ubuntu ships that by default.
                    texture-float was passed when building Mesa and S2TC is present by default in Ubuntu (sure wish there was a way to automatically parse the Mesa build configuration of the resulting binaries; similar to gcc -v, so that the information could be presented...)
                    Michael Larabel
                    http://www.michaellarabel.com/

                    Comment


                    • #11
                      These are really good results!

                      In many benchmarks, the 70-90% target has been reached, which is fantastic. There is a lag with ridiculously high FPS, which is expected and irrelevant, and in some games (Doom3, RQuake). But these games should be examined so specific bottlenecks causing the performance drops can be found and addressed.

                      With UVD already here, GL3.3 around the corner, good performance, OpenCL progress and (hopefully) proper power management coming soon, this could be an amazing year for AMD users!

                      Comment


                      • #12
                        Originally posted by GreatEmerald View Post
                        But that's what the freedesktop wiki is for.
                        It's nearly there, but according to this:

                        http://www.x.org/wiki/RadeonFeature#...Radeon_Drivers

                        Radeon power management looks like done when its only static PM I suppose, and HDMI audio is marked as DONE for several generations, while AFAIK its not really on par with the windows driver yet, so something more subtle may be missing.

                        Comment


                        • #13
                          Awesome work, congratulation to the developers!

                          Maybe by the time the OSS drivers are able to fully support Steam games (OpenGL wise) the OSS radeon drivers will be close enough to finally forget about catalyst and then i can upgrade to xorg-server 1.14. I still have a feeling there's a lot of performance optimizations left unfound or unpolished, but it's really awesome to see that it doing GL2 games running so well now. I'll have to try it in a few months (when this stuff is on Arch) and see how well Blender performs under the OSS drivers.

                          Comment


                          • #14
                            It would be nice if there was benchmark that could put drivers on their knees, for example....Heaven.

                            Comment


                            • #15
                              Originally posted by pingufunkybeat View Post
                              These are really good results!

                              In many benchmarks, the 70-90% target has been reached, which is fantastic. There is a lag with ridiculously high FPS, which is expected and irrelevant, and in some games (Doom3, RQuake). But these games should be examined so specific bottlenecks causing the performance drops can be found and addressed.

                              With UVD already here, GL3.3 around the corner, good performance, OpenCL progress and (hopefully) proper power management coming soon, this could be an amazing year for AMD users!
                              3.2 is more important. OSX use 3.2 excessively, and apps on Win are using DX (more likely). So people will tend to have 3.2 code paths for the sake of OSX compa. Games also should pick 3.2 (as this give 95% of Steam market...), while 4.x will still for some time be add-on, as it require up to date drivers on win, mean no osx support, and binarys on Lin.

                              Yes 3.3 will mark "full" OpenGL support for class-DX11 hw, but otherwise 3.2 support is more important (and we will get them both in same mesa release at least for intel & r600g)

                              Comment

                              Working...
                              X