Announcement

Collapse
No announcement yet.

Ubuntu 23.04 Laptop Performance Mixed Against Ubuntu 22.10

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

  • #21
    Wow, this is a joke of a release!
    They didn't even bother to run one the Phoronix benchmarking tool?
    I guess they were too busy working only on the Snap crap that they are still trying to push...

    Comment


    • #22
      I browsed around gnome gitlab and found 3 bug reports related to problems with programs going fullscreen:

      https://gitlab.gnome.org/GNOME/mutter/-/issues/2712 [GpuTest benchmarks fail to run fullscreen]
      https://gitlab.gnome.org/GNOME/mutter/-/issues/2639 [44.rc: Metro Exodus fullscreen mode slightly misplaced, does not fill screen completly]
      https://gitlab.gnome.org/GNOME/mutter/-/issues/2678 [44.rc: breaks Fullscreen mode on most SDL based games (Ren'Py, Unity, FNA, Source-Engine, ...) running through X/Xwayland]

      Comment


      • #23
        That's why I stick to Debian, Ubuntu is in perpetual beta.

        Comment


        • #24
          Originally posted by hax0r View Post
          The 3D/game performance drop in 23.04 appears to be gnome-shell/mutter related, at least to me on i7-4970K & GTX 980. I switched to openbox session on 23.04 and got all the performance back, and more.

          On 23.04 the glmark2 appears to have a problem going fullscreen (the window's headerbar stays visible), you can try it for yourself:
          Code:
          glmark2 --annotate -s 2560x1440 --fullscreen
          I'm speculating the glmark2 (and many other programs) are not undergoing 'fullscreen unredirect' (not bypassing compositor) leading to much lower performance.

          I was also seeing this in journalctl:
          Code:
          Apr 20 21:38:25 ROG gnome-shell[1967]: Can't update stage views actor <unnamed>[<MetaWindowActorX11>:0x5558f10ec530] is on because it needs an allocation.
          Apr 20 21:38:25 ROG gnome-shell[1967]: Can't update stage views actor <unnamed>[<MetaSurfaceActorX11>:0x5558f49f9a30] is on because it needs an allocation.​


          Reminder: all the 'blue' bars in graphs below belong to my system:
          - i7-4970K with mitigations=off and cpupower set to 'performance', X11 display server, GTX 980 on nvidia 525 driver ​​

          Hey, nice job on finding out the cause of the regression!
          Keep it up, we need more people like you around here!

          In general, I can highly recommend OpenBox for any kind of gaming, since it is so lightweight that it basically doesn't take away any resources on your system.

          Comment


          • #25
            Looks like it's indeed a mutter 44 issue and not specifically Ubuntu:

            On Ubuntu 23.04 the glmark2 performance is much lower than in 22.10, glmark2 (as well as many other 3D/games) have problems going fullscreen (the window's headerbar stays visible),...

            This takes inspiration from FVWM and implements double parenting of decorated windows, although perhaps for different reasons. When decorated, we'll now have ...

            Comment


            • #26
              Originally posted by hax0r View Post
              You sure about this? What's your evidence?
              If I was sure I wouldn't have said "I wonder"

              Comment


              • #27
                Originally posted by royce View Post
                Looks like it's indeed a mutter 44 issue and not specifically Ubuntu:

                https://gitlab.gnome.org/GNOME/mutter/-/issues/2770
                Yes that was me, figured I would submit a bug and share some of my findings with the devs instead of complaininig about state of GNU/Linux on desktop on phoronix.

                tl;dr, did some more testing today, I patched ubuntu's mutter with 2941 and 2961 merge requests, the https://gitlab.gnome.org/GNOME/mutte..._requests/2961 did fix the problem with glmark2 not being able to enter fullscreen and performance is on par with 22.10, however in windowed mode the huge regression is still there (not to mention problems that 2941 MR brings such as games starting up minimized/unvisible but audio still continues playing in background etc):


                Comment


                • #28
                  Originally posted by hax0r View Post
                  Yes that was me, figured I would submit a bug and share some of my findings with the devs instead of complaininig about state of GNU/Linux on desktop on phoronix.
                  I guess in those GLMark2 numbers KDE and Gnome instances are all Wayland and the openbox wm is X11 bare metal.

                  Down a little in performance due to Wayland compositor/tear free that understandable due to different AMD driver issues(a little 10-20%).
                  I did one last measurement with AMDVLK to verify if X also benefited from the reduced swapchain images, and sure it did! Here is my (hopefully final) table: 95ms (5.7 frames) on Composited X (RADV) 87ms (5.3 frames) on Composited X (AMDVLK) 77ms (4.6 frames) on Kwin Wayland (RADV) 75ms (4.5 frames) on Weston (RADV) 65ms (3.9 frames) on Uncomposited X (RADV) 54ms (3.3 frames) on Uncomposited X (AMDVLK) 54ms (3.3 frames) on Kwin Wayland (AMDVLK) 53ms (3.2 frames) on KMS (RADV) 51ms (3.0 frames)...

                  Down over 50% that not right if the compositor is Wayland. X11 compositor disappearing 50% on low end hardware can be status normal.

                  Comment

                  Working...
                  X