Announcement

Collapse
No announcement yet.

Ubuntu 13.04 Desktop Gaming Performance Comparison

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

  • #16
    Don't composite fullscreen windows

    I guess that all compositor where tested without their respective option to disable compositing on fullscreen. That would explain why they are all in the same ballpark. For Enlightenment, we don't turn this feature on by default as most driver out there are buggy and I guess all compositor have the same behavior. So I recommend to go in Settings/Composite/Advanced/Memory/Don't composite fullscreen windows for Enlightenment if you want better number.

    Now if you look at the number as a benchmark for just how fast compositor are at pushing frame, it is still interesting. I am wondering why some test case show E17 as been slow when other it is more at lead of other composite manager. I should spend time playing I guess :-)

    Comment


    • #17
      Slightly off topic but Michael weren't you supposed to be publishing some steam benchmarks? Are these comming?

      Comment


      • #18
        Originally posted by mark_ View Post
        so: the less complex the WM the more FPS. What about xdm? If Xfce gives me 10fps more than KDE and xdm gives me 10fps more than Xfce this could be totally worth it for a gaming machine. Please include xdm next time.
        I think you are after twm. xdm is a display manager and has nothing to do with window management.

        Comment


        • #19
          Enlightenment e17 was benchmarked with stock settings, which does mean it does not redirect fullscreen windows. It automatically does this, if the option is turned on. For full compatibility it is not default to on.

          Comment


          • #20
            Originally posted by mark_ View Post
            so: the less complex the WM the more FPS. What about xdm? If Xfce gives me 10fps more than KDE and xdm gives me 10fps more than Xfce this could be totally worth it for a gaming machine. Please include xdm next time.
            I doubt that's really the reason. Most WMs *shouldn't* really be executing much of anything when a App is fullscreen. My guess is that these results reflect Windowed Game performance with default WM settings. Obviously OpenBox & Xfce are going to win here, since they're non-composting. That's just a guess though, we need more info to know for sure.


            Originally posted by oleid
            One thing that I find interresting ist, that gnome-shell is sometimes the slowest and sometimes (one of) the fastest desktop(s) according to these benchmarks. How can that be? Shouldn't there be a uniform slowdown due to composite?
            It's possible some games where run in fullscreen while others where not. We need more info. It's also possible (if the games where run in fullscreen) that specific games don't properly request fullscreen, and therefor Gnome/Unity/KDE composite them anyways. I don't know for sure, but it is interesting.

            Comment


            • #21
              Gala?

              I'd really like to see the performance of Gala, the coming-up mutter-inspired window manager of Elementary OS. It gives a good user experience but I'm really curious of the true capabilities.

              Comment


              • #22
                Originally posted by Ericg View Post
                Michael, clarify here in the forums: was KDE set to unredirect fullscreen windows or not? If so, great, then obviousl theres a regression. But If not then this entire article can be tossed out because the basic premise is "Window manager doesnt matter."

                Like I really dont bash your articles much but you not setting that option from time to time is mine--and others-- biggest peeve with you because theres a certain line between "Keeping the defaults" and "common sense for benchmarking."
                I disagree, sticking to the default is the most realistic test. What is 'unredirect fullscreen'?
                Every benchmark someone says it's useless because: "Distro Y sucks, distro X wasn't tested", "gcc didn't A,B,C parameters", "kernel ....", "Testing version 3.5.71 is stupid", "Gosh, proper gaming is done with low latency real time kernel."....

                Last time I saw a blog post by kwin dev about Phoronix benchmark. He thought the benchmark was useless, because it didn't use kwin with three different drawing back-ends and n other settings. If everyone had their way, each chart would have thousands of items.

                Comment


                • #23
                  Originally posted by talvik View Post
                  I disagree, sticking to the default is the most realistic test. What is 'unredirect fullscreen'?
                  Every benchmark someone says it's useless because: "Distro Y sucks, distro X wasn't tested", "gcc didn't A,B,C parameters", "kernel ....", "Testing version 3.5.71 is stupid", "Gosh, proper gaming is done with low latency real time kernel."....

                  Last time I saw a blog post by kwin dev about Phoronix benchmark. He thought the benchmark was useless, because it didn't use kwin with three different drawing back-ends and n other settings. If everyone had their way, each chart would have thousands of items.
                  If you do gaming benchmarks, do them right.

                  These benchmarks are an absolute joke. What's the point of these? Have you got any conclusion at all? Do you even know the conditions these were done on?

                  I don't understand "sticking to the default is the most realistic test" either. What's the most realistic for you? Should all distros using openbox by default (as it appears to be the best for gaming)? AFAIK, in KDE you just have to press 'Shift+Alt+F12' to disable compositing or check the option 'Disable compositing in fullscreen' and do the tests on fullscreen.

                  Comment


                  • #24
                    Originally posted by talvik View Post
                    I disagree, sticking to the default is the most realistic test. What is 'unredirect fullscreen'?
                    Every benchmark someone says it's useless because: "Distro Y sucks, distro X wasn't tested", "gcc didn't A,B,C parameters", "kernel ....", "Testing version 3.5.71 is stupid", "Gosh, proper gaming is done with low latency real time kernel."....

                    Last time I saw a blog post by kwin dev about Phoronix benchmark. He thought the benchmark was useless, because it didn't use kwin with three different drawing back-ends and n other settings. If everyone had their way, each chart would have thousands of items.
                    Even on Windows you have to manually install drivers to get good gaming performances. Mom/Grandma types aren't going to understand how to do that, but Gamers are. The same rules apply to Linux gamers really. If everywhere on forums/blogs/etc people are saying "check this setting in your DE for better game performance" then a lot of folks are going to do it.

                    Comment


                    • #25
                      Benchmarks don't show anything hugely surprising; the lightest desktops/non-composited WM's will yield best performance.

                      I too am curious how KDE is used though and whether there is a regression. Either way shouldn't affect me too much, I just disable desktop effects before i play a game anyway since KDE makes it so easy to do (toggle with alt-shift-F12)

                      Comment


                      • #26
                        What I notice about these tests is that excluding a few worst cases, even composition isn't as bad as it used to be to game. The DE and driver makes have really cleaned up thier work to prevent slow down alot and it shows.

                        Comment


                        • #27
                          Here is the result of test I did some time back

                          Link

                          Results of fgl_glxgears (5 sec)

                          Gnome Classic (without effect): 4540.2
                          Ubuntu Unity: 3935
                          Gnome Classic (with effect): 3881.6
                          KDE Plasma Workspace: 3776.8
                          Gnome: 3693.2
                          Cinnamon 3D: 3679.4
                          Ubuntu (Unity 2D): 1307
                          Cinnamon 2D: 1104

                          Comment


                          • #28
                            Originally posted by xpressrazor View Post
                            Ubuntu (Unity 2D): 1307
                            Cinnamon 2D: 1104
                            That's odd... Maybe because they use something fglrx doesn't support? wow...

                            Comment


                            • #29
                              There were two things I got from those benchmarks: 1) light-weight desktops are in general slightly faster than heavyweights (but usually not more than about 20% which implies for casual games it may not be an issue), and 2) there are a few unusual cases like Prey on Unity where the slowdown is higher than expected.

                              Michael, you're benchmark articles tend to lack two things: 1) a conclusion (like the above, expanded a little but still short, at the end) and 2) a little speculative discussion on why, e.g. one of the first questions in the forums was:
                              Michael, clarify here in the forums: was KDE set to unredirect fullscreen windows or not?
                              Especially since this feature has to be selected manually but has a significant effect it's worth a quick mention.

                              Comment


                              • #30
                                we need a universal call, like Windows has, for programs to be able to tell the WM to disable compositing when needed.

                                Comment

                                Working...
                                X