Announcement

Collapse
No announcement yet.

Radeon Gallium3D Still Long Shot From Catalyst

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

  • Radeon Gallium3D Still Long Shot From Catalyst

    Phoronix: Radeon Gallium3D Still Long Shot From Catalyst

    Following recent advancements in the open-source Radeon Linux driver like 2D color tiling support, I've carried out some new benchmarks of the open-source Radeon Gallium3D driver compared to AMD's official Catalyst driver. This time around, the open-source driver is seeing tests against AMD's binary blob when various performance-optimizing tweaks are enabled to see where the performance stands today.

    http://www.phoronix.com/vr.php?view=17178

  • #2
    What a surprise...

    Comment


    • #3
      Thanks for the updated comparison Michael!

      I know I am maybe asking too much. But I think it would be quite interesting to see traces(apitrace supports CPU profiling) of the benchmarks for the very slow results at least, just to get a better idea where the binary drivers are faster.

      Btw I think one could actually use api traces of games as benchmarks. This would additionally ensure that the same call paths are executed, no fall-backs or workarounds for specific hardware taken.

      PS:
      Can someone point me to a tutorial of some sort on how to get the latest mesa stuff running next to catalyst. Are phoronix test setups documented?

      Comment


      • #4
        Michael, where is the Doom3 test?
        I still consider this wide gap between r600g and Catalyst on higher res/quality to be due to still missing optimizing shader compiler.
        At lower res/quality GPU can keep with unoptimized shaders, but with lots of pixels the game changes. What happened with Vadim Girlin's initiative on that?

        nvidia had hardware scheduler so far, and that is the reason for better nouveau speed, event unsupported by nvidia, but now with kepler, they go AMD route, with software scheduler in the driver ( at least thats what anandtech says ).

        Comment


        • #5
          Originally posted by log0 View Post
          Can someone point me to a tutorial of some sort on how to get the latest mesa stuff running next to catalyst. Are phoronix test setups documented?
          Three choices:
          * Use a distribution like gentoo that actually allows several libGL implementations to exist on your system. You need to load modules manually, and switch between two xorg.conf (i.e. forget XDM), but it can work.
          * Just install a separate OS on a different partition. Maybe play around with UnionFS to save space.
          * Try to manually install both on an unsupported system, go crazy over the conflicts that arise, eventually bork your system and reinstall.

          Honestly, if you have to ask how to do it, go with option 2. It's safer.

          Comment


          • #6
            Debian has glx-alternative packages for allowing binary/open-source GPU drivers to coexist.

            Comment


            • #7
              Yeah, catalyst sure does have a long way to go before it has the utility and openness of radeon-g3d.

              Comment


              • #8
                Originally posted by droidhacker View Post
                Yeah, catalyst sure does have a long way to go before it has the utility and openness of radeon-g3d.
                LOL you always can buy a faster card but you can't buy a open catalyst

                Comment


                • #9
                  I don't see it so doomy and gloomy.

                  In trivial pixel-pushing benchmarks like OpenArena, Warsow and Padman, the OSS drivers are slower, but well within playable range. 500+ FPS is really not anything anyone cares about. 120Hz should be enough for anyone :P

                  With a more complex game at high resolution (Xonotic @1080p), the OSS drivers are at 60-65% of performance. This is a really good number. With HiZ and some shader compiler optimisation, it should reach 75%, which is really close to the best that can be reasonably expected from the open stack. At 75% performance, the OSS drivers becomes a real option for everyone other than people who need the absolute ultimate performance (scientific computing and elite gamerz). It's fast enough for the desktop and casual gaming, and the missing performance costs 20 bucks if you really need it. That would be wonderful! And don't forget that future generations (HD 7000) should be easier to optimise due to the switch away from VLIW.

                  It's interesting to see that Nexuiz is so much slower than Xonotic, as they are basically the same game... No idea what happened there. Also, more challenging benchmarks like doom3 would be nice.

                  Comment


                  • #10
                    I really don't understand why there are no doom3 benchmarks: xonotic and doom3 were the only two benchmarks I was interested to see.
                    ## VGA ##
                    AMD: X1950XTX, HD3870, HD5870
                    Intel: GMA45, HD3000 (Core i5 2500K)

                    Comment


                    • #11
                      Originally posted by pingufunkybeat View Post
                      With a more complex game at high resolution (Xonotic @1080p), the OSS drivers are at 60-65% of performance. This is a really good number. With HiZ and some shader compiler optimisation, it should reach 75%, which is really close to the best that can be reasonably expected from the open stack. At 75% performance, the OSS drivers becomes a real option for everyone
                      I would have agreed if we were nearing 75% of catalyst in Unigine Heaven.
                      ## VGA ##
                      AMD: X1950XTX, HD3870, HD5870
                      Intel: GMA45, HD3000 (Core i5 2500K)

                      Comment


                      • #12
                        Originally posted by log0 View Post
                        Btw I think one could actually use api traces of games as benchmarks. This would additionally ensure that the same call paths are executed, no fall-backs or workarounds for specific hardware taken.
                        Regrettably, this doesn't work. When you profile an apitrace replay, you find that a huge portion of the profile is simply apitrace parsing the multigigabyte trace file.

                        Comment


                        • #13
                          Originally posted by Qaridarium View Post
                          LOL you always can buy a faster card but you can't buy a open catalyst
                          Sure you can, it just costs about $10 bazallion.

                          Comment


                          • #14
                            I think it's very promising. The Xonontic benchmarks are very pleasing.

                            I am guessing from the benchmarks is that there is still some stuff falling back to software that is killing performance for certain things. With some optimization to applications and filling in some missing pieces in the drivers and we are golden. Once open source gets within about 70-80% of proprietary then I'd call it success.

                            Comment


                            • #15
                              Originally posted by mattst88 View Post
                              Regrettably, this doesn't work. When you profile an apitrace replay, you find that a huge portion of the profile is simply apitrace parsing the multigigabyte trace file.
                              Hmm, I've got a couple traces from games and my own stuff(20-70fps, 100-400MB). And they take about the same time.

                              Just did a quick run with vdrift about 2min, 130MB trace. Frame rate without tracing is about 22fps, with tracing 17fps, retracing 15fps(68% of original fps). Are my results atypical?

                              As I see it, the slowdown would be the same for all benchmarked cards and we are interested in the relative performance only.

                              Comment

                              Working...
                              X