Announcement

Collapse
No announcement yet.

Gaming/Graphics Performance On Unity, GNOME, KDE, Xfce

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

  • Gaming/Graphics Performance On Unity, GNOME, KDE, Xfce

    Phoronix: Gaming/Graphics Performance On Unity, GNOME, KDE, Xfce

    It is going on a year since showing how Unity, Compiz, GNOME Shell & KWin affect graphics/gaming performance, so here is an updated 2012 look. In this article are a variety of OpenGL benchmarks run under the current latest desktops as will be found in Ubuntu 12.04 LTS: Unity, Unity 2D, GNOME Shell, GNOME Classic, KDE Plasma, and Xfce. AMD and NVIDIA graphics were tested with both the latest closed and open-source drivers.

    http://www.phoronix.com/vr.php?view=17046

  • #2
    The compiz version in Ubuntu 12.04 is still the old 5 month old build from September 2011, that is also included with Ubuntu 11.10. The new release with about 10 or so important performance fixes from Daniel Van Vugt and more should be included shortly before Ubuntu 12.04 Beta 1. Also, some remaining Unity performance issues also have patches ready to go upstream very shortly. So the Unity (+ Compiz) Performance will go up quite a bit before Ubuntu 12.04 releases.

    Comment


    • #3
      @author of benchmarks
      please don't use similar colors! really.

      I have a hard time distinguishing between Unity2D and Xfce cause they're both dark red.

      Comment


      • #4
        Wow, Gnome Shell is doing pretty damn good. Looks like it's second only to XFCE in most of the tests, which I kind of expected. I expected that Gnome Classic would be faster than Shell, but surprisingly not.

        Comment


        • #5
          KDE with or without compositing? I guess, with?

          Comment


          • #6
            Originally posted by benmoran View Post
            Wow, Gnome Shell is doing pretty damn good. Looks like it's second only to XFCE in most of the tests, which I kind of expected. I expected that Gnome Classic would be faster than Shell, but surprisingly not.
            It used to be pretty damn awful, so that is a good thing. Also, I think they tested Gnome Classic WITH compositioning rather than without.

            Comment


            • #7
              Very good comparison. Since a few weeks I asked myself how to modify up my low-cost system so that it runs World of Padman as fast as possible. Now I know.
              Thank you!

              Comment


              • #8
                Why isn't LXDE considered? My only guess is that it's not complete enough on it's own.

                Comment


                • #9
                  How can performance under catalyst be so uniform?

                  There must be some external reason that the performance under catalyst is that uniform across so many tests and so many different desktops. I think it must be that the limitation is some vsync setting, or some other limit other than the performance of those desktops. Remove that limit, and I think the results will change a lot.

                  Comment


                  • #10
                    Originally posted by Bruners View Post
                    Why isn't LXDE considered? My only guess is that it's not complete enough on it's own.
                    i was wondering the same thing. personally i like lxde (i'm using it right now) and i found it to be noticeably faster than xfce. i personally hate xfce, i'd rather use unity. back before the gnome 3 days, xfce was just about as heavy as gnome 2, but offered less features and yet it strived to be lightweight.

                    lxde is the only true lightweight DE. sure its missing a lot of features but i like that because it doesn't force you to install a ton of gnome or kde packages and it only runs the minimal.

                    Comment


                    • #11
                      One big caveat with GNOME Shell is that it does only unredirect fullscreen SDL windows, which covers a lot of native games, but not anything running in Wine.

                      I have seen a few cases where performance in a game drops to the floor because of this, so for the moment I'm patching Mutter to unredirect anything running fullscreen:

                      Code:
                      Enable unredirect for all fullscreen windows. Not just unredirect-override.
                      
                      --- mutter-3.2.1.orig/src/compositor/meta-window-actor.c
                      +++ mutter-3.2.1/src/compositor/meta-window-actor.c
                      @@ -1219,7 +1219,7 @@ meta_window_actor_should_unredirect (Met
                         MetaScreen *screen = meta_window_get_screen (metaWindow);
                         MetaWindowActorPrivate *priv = self->priv;
                       
                      -  if (meta_window_is_override_redirect (metaWindow) && priv->opacity == 0xff && !priv->argb32)
                      +  if (priv->opacity == 0xff && !priv->argb32)
                           {
                             int screen_width, screen_height;
                             MetaRectangle window_rect;

                      Comment


                      • #12
                        Originally posted by Bruners View Post
                        Why isn't LXDE considered? My only guess is that it's not complete enough on it's own.
                        I agree, and performance on Fluxbox would be interesting as well.

                        Comment


                        • #13
                          Gnome shell and ATI Did it work well?

                          I missed enlightment E17 and Pantheon alpha, Mate, openbox, razor qt, and fluxbox.

                          But I have an ATi HD 4250 and i have not been able to run Gnome Shell or Cinnamon with catalyst drivers because a bug in Mutter / ATI

                          How did you solve this bug?

                          Is it bypassed anyway?

                          I do like to have installed 3 distros / experimental now sabayon + backup LMDE + main Mint12.

                          And test every desktop.

                          1 consideration Ubuntu Gnome is not true Gnome - at least I have just read the only way of a "true" gnome in ubuntu is installing in text mode the server version and from the console install Gnome from Gnome PPA

                          2 If there are a way of use Gnome Shell with my integrated ATI HD 4250 I would like to know the trick. i also have a Nvidia 9600, but I do not want to plug in it only to test Gnome 3.

                          3. Perhaps in other post, E17, Pantheon, Openbox alone and +KDE and +Gnome, Mate, Cinnamon, razor qt and fluxbox, as alternative desktops - Or any other one i miss - would be nice to be tested.

                          4.- As at any good article, as i use to read here, i do like to read the 5W, and I would like to read something about why the Catalyst driver is almost independent from the desktop in gaming, and Nvidia graphic drivers aren't. also I would like to know this tests in other OSs as MS WOS and OSX - perhaps it will never happen at phoronix, i do not know why - and some explanation of why the difference of performance between one drivers or others. In this case the lack of standard deviation in AMD drivers I do not know if is better or worse because it is a maximum for all or a minimum for all, what I can imagine more if I have the same tests with other OSs.

                          Thanks in advance, I enjoyed a lot reading this article, but I always have more questions when I read something like this. And if you do not answer this nobody else will do it at my feeds. Because you are my best Linux benchmark blog, and perhaps the best over the net.
                          Last edited by mitcoes; 02-08-2012, 11:54 AM.

                          Comment


                          • #14
                            Originally posted by mitcoes View Post
                            I missed enlightment E17 and Pantheon alpha, Mate, openbox, razor qt, and fluxbox.
                            Some of those are just window managers which work inside Gnome/KDE and can't be taken into consideration in a test like this, for razor qt it is still in early developement.

                            Comment


                            • #15
                              reaction from the kwin developer:

                              http://blog.martin-graesslin.com/blo...g-performance/

                              Comment

                              Working...
                              X