Announcement

Collapse
No announcement yet.

Does Compiz Still Slow Down Your System?

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

  • #16
    While we are talking about performance impacts from composited environment, I have a question for all of you: do you have problem with watching a VDPAU accelerated video while composition is turned on?

    I have such a problem on my GTX 470. While having kwin composition enabled, if I open a movie in SMplayer (or mplayer) configured to use VDPAU, X freezes after a while, or after skipping video (jumping to the middle of movie or similar), or sometimes immediately after starting video playback. This happens always while watching 720p or higher resolution videos, and sometimes even with smaller videos. Although X is freezed, key events are processed so I am able to pause playback (this unfreezes the X).

    If I disable composition (suspend desktop effects), playback runs normally.

    The system is Arch x64, with KDE 4.6, nvidia gtx 470, and default driver from Arch repository (260.19.44-1).

    Comment


    • #17
      Originally posted by Drago View Post
      Is it just me, or Radeon FOSS drivers are some like %50 of Catalyst. I don't know about lower resolutions, but at HD things are fantastic
      That was also on my mind. On some tests it was even higher than that. Not bad at all.
      Also, KWin tests as well would be great!

      Comment


      • #18
        Originally posted by marek View Post
        I have noticed too that Compiz mostly doesn't decrease performance. However, KWin does and the difference can be huge.
        Uh? I did lots of benchmarks with Gallium and Kwin-4.6 enabled/disabled and there was no difference at all when using full screen mode.
        ## VGA ##
        AMD: X1950XTX, HD3870, HD5870
        Intel: GMA45, HD3000 (Core i5 2500K)

        Comment


        • #19
          Compiz sucks for gaming

          For me Compiz sucks when gaming both on Nvidia and Ati. Killing Compiz manually when entering game is such party breaker so I made scrips to do this automatically when entering game, and revert this when exiting game. Here are the scripts, hope someone finds them useful:

          http://www.techytalk.info/2010/06/co...witch-scripts/

          Cheers

          Comment


          • #20
            Originally posted by darkbasic View Post
            Uh? I did lots of benchmarks with Gallium and Kwin-4.6 enabled/disabled and there was no difference at all when using full screen mode.
            KWin always disables compositing for full-screen apps (like deanjo says, there is an option controlling this, and many distros use it by default).

            Comment


            • #21
              Originally posted by Raine View Post
              Looking at those results, mostly on Nvidia side, makes me wonder how much efforts are being dedicated by Gallium3D team to improve performance on Gallium3D infraestructure.

              Does anyone know if there is any work beeing done to improve Gallium3D infraestructure performance in the wild?

              I didn't notice any special/cool patch for Gallium3D on 2.6.38. I could be wrong, and wish to Please correct me if i'm wrong....
              How much effort in the Gallium3D architecture? As far as I know, not much. Work is happening on the gallium drivers.

              But there are outstanding patches that entered the 2.6.38 (new memory manager, pageflip, zcompression) AND mesa (merge of the nvc0 code into the nv50) lately and that brought a really nice performance improvement

              Comment


              • #22
                Yes, it does. Nvidia Quadro 4600 fx.

                Comment


                • #23
                  havent noticed any perf drops with compiz on.
                  ati 5770 user, with catalyst drivers

                  Comment


                  • #24
                    Originally posted by xpander View Post
                    havent noticed any perf drops with compiz on.
                    And there is a statement that really matters. As long as the end user does not perceive any noticeable obstructions to using their software, then it really doesn't matter. The real issue isn't if it drops performance but if it drops performance to a point where disabling compiz becomes a must to run applications effectively.

                    Comment


                    • #25
                      This is an interesting benchmark and I should check that on my machine.
                      There seems to be a huge performance drop with compiz with the Nvidia binary driver, whic was totally unexpected ... I always considered gaming performance with compiz to be on par with metacity when using Nvidia's driver, which may not be true anymore.

                      The last game I've benchmarked, Left4dead 2 on Ubuntu and Windows 7, did not show any performance drop with Compiz, but then again it was with 260.x drivers.

                      Phoronix articles benchmarks games based on old technologies, so when benchmarking on my machine, I'll make sure to add a few newer games.

                      I'll be running OpenArena, World of Padman, Nexuiz, and Urban Terror to compare with phoronix results but I can't publish a benchmark with 3 games based on Quake 3 Arena's engine and one game based on Quake 1 ...
                      I'll add to the mix the following :

                      - UT 2004
                      - Doom 3 or Quake 4 or both
                      - Unigine benchmark
                      - Half Life 2 Lost Coast, with latest Wine version
                      - Maybe Oblivion, if I find a way to benchmark it. (still with Wine)
                      - An Unreal Engine 3 game would be cool too. (still with Wine)
                      - Maybe more ...

                      Finally, I'll implement WM switching in Lutris ( http://lutris.net ), the options have been in the GUI since the start of the project but hasn't been doing anything ...

                      Comment


                      • #26
                        Originally posted by Strider View Post
                        Phoronix articles benchmarks games based on old technologies, so when benchmarking on my machine, I'll make sure to add a few newer games.

                        I'll be running OpenArena, World of Padman, Nexuiz, and Urban Terror to compare with phoronix results but I can't publish a benchmark with 3 games based on Quake 3 Arena's engine and one game based on Quake 1 ...
                        Because the open-source drivers can't handle Unigine or anything more interesting.
                        Michael Larabel
                        http://www.michaellarabel.com/

                        Comment


                        • #27
                          They can handle much more complicated stuff than Quake3, though:

                          Doom3
                          Quake4
                          Penumbra games
                          Amnesia

                          These are not open source, but that's an issue with open source games, and not open source drivers.

                          Comment


                          • #28
                            Originally posted by pingufunkybeat View Post
                            They can handle much more complicated stuff than Quake3, though:

                            Doom3
                            Quake4
                            Not well without S3TC.

                            Originally posted by pingufunkybeat View Post
                            Penumbra games
                            Amnesia
                            I see no automated tests or test profiles.
                            Michael Larabel
                            http://www.michaellarabel.com/

                            Comment


                            • #29
                              I finished Doom3 and RoE without S3TC.

                              Quake4 requires S3TC, but is this a problem? That works now, the game runs on Radeon, and probably also on other drivers.

                              Amnesia and Penumbra might be difficult to benchmark automatically, you're right.

                              Comment


                              • #30
                                Originally posted by Michael View Post
                                Because the open-source drivers can't handle Unigine or anything more interesting.
                                That's correct, I concluded the same thing when I tried benchmarking Nouveau. It's still possible to benchmark UT2004 though.

                                I'll try to go as far as I can with the Nouveau driver but my main focus is on NVidia proprietary driver.

                                While I'm at it, I'll also compare performance drop with and without Unity, even if it still has serious issues that make it inappropriate for gaming (the panel does not hide when switching to full screen games). I've noticed unusual performance drop with Unity which shouldn't happen since it's only a Compiz plugin.

                                Originally posted by pingufunkybeat
                                I finished Doom3 and RoE without S3TC.
                                Strange, I couldn't get Doom 3 to run with Nouveau, I'll check on that again.

                                Comment

                                Working...
                                X