Intel Graphics On Ubuntu: GNOME vs. KDE vs. Xfce vs. Unity vs. LXDE

[phoronix]

Phoronix: Intel Graphics On Ubuntu: GNOME vs. KDE vs. Xfce vs. Unity vs. LXDE

For those wondering how the Intel (U)HD Graphics compare for games and other graphical benchmarks between desktop environments in 2018, here are some fresh benchmarks using GNOME Shell on X.Org/Wayland, KDE Plasma 5, Xfce, Unity 7, and LXDE.

http://www.phoronix.com/vr.php?view=25901

[pracedru]

Those are weird results on Unigine Value, where fullscreen is much slower than windowed.

[raonlinux]

Thanks Michael for the Benchmarks, it wil be nice to try another distro as Fedora, that is the main that you use.

[Goddard]

I think it would be interesting to also see the Deepin desktop and Budgie.

[Kayden]

Those are weird results on Unigine Value, where fullscreen is much slower than windowed.

We've definitely observed bizarre results with Unigine as well, where Valley and Heaven have wildly different performance in windowed vs. fullscreen mode. I'm not sure anyone's figured it out...

[slacka]

What feature is Wayland missing that's keeping it from performing as well as X11?

[treba]

Thanks a lot for these benchmarks, Michael ! So there's still a serious performance penalty on using gnome-wayland in certain conditions. That's a bit unfortunate, as fedora and ubuntu 18.04 both use it as the default session (which is great concerning most other aspects IMO).

I wonder if the difference is because of the missing fullscreen bypass or rather an issue about xwayland. The benchmark for unigine valley in windowed mode would indicate the later. For the first one, there's still a open ticket which I hope daniel or jadahl will have time tackling before gnome 3.28 gets released. But I guess it will have to wait till 3.30.

[droste]

We've definitely observed bizarre results with Unigine as well, where Valley and Heaven have wildly different performance in windowed vs. fullscreen mode. I'm not sure anyone's figured it out...

By the looks of it, I would say that as soon as they go fullscreen, compositing is turned off (which I think all of the window managers do these days) and they run rendering to the screen buffer with vsync turned on.
10fps = 60Hz / 6
14 fps = 60Hz / 4

This is not a problem when the app is rendering to their own off screen buffer in the composited case.
I could be wrong though, it's just a guess from the numbers.

/edit:
Would also explain why there's virtually no SE

[Happy Heyoka]

KDE Plasma on Wayland wasn't tested since when installing the Plasma Wayland session package on Bionic it would crash right away when trying to log-in.

I had a fun day with that on Friday... something in the current Mesa Git makes KDE very unhappy deep down in the Qt stuff (some kind of threading of shaders issue?).
Purging the Mesa Git PPA brought me back and I've been too busy to poke around and do a bug report for it.

[SpyroRyder]

What feature is Wayland missing that's keeping it from performing as well as X11?

It's probably more that most games will currently under wayland use XWayland and thus have a bit of an overhead. We unfortunately wont see that improve for most currently released games, and even with some unreleased games as they target x11 due to that being what Ubuntu 14.04 and 16.04 shipped. We'd probably have better support now had Ubuntu gone with Wayland from the start but then again games companies even on Windows often target older software due to the time it takes to develop large games.