KDE/GNOME Wayland vs. X.Org Radeon Linux Gaming Performance

[aufkrawall]

...those are completely different benchmarks, and the results don't contradict anything said here or there. The Wayland sessions for KDE and GNOME both appear to outperform their X counterparts.

With active window compositing. There is ~0 rendering performance difference between Xorg with window manager compositing off and Wayland with direct scanout (which currently works more often with Mutter than with KWin, but this should about to change soon) on dGPUs.
If Michael configured his benchmark environment so that KWin compositing would not be automatically enabled on log-in, the Xorg results likely would look different.

[tuxd3v]

well on these, we saw X kicking, but it was mostly with productivity workloads..
https://www.phoronix.com/scan.php?pa...-Wayland-Power

It would be nice to have the same power consumption and temperature graphics( on this bench), so that we could have a picture on them..

Now, I was left with the impression that Xorg works nicely for productivity and Wayland works better for Games o.0

[SkyWarrior]

Gnome Wayland is faster but meaningless if I cannot use Freesync 144hz on my monitor. Until then X.Org it is for gaming.

[ernstp]

Wayland is not only faster, but more secure and power efficient. Eat this xorg fanboys.

It moves quickly so if you haven't tried it in a year it might be quite different.

[damentz]

I hope a kwin developer takes a look at these graphs. It appears that KDE always scores lower, even if it looks margin of error. Most likely there's some weird frame timing bug or some strange interaction between the compositor and the game being composed.

Especially with Hitman 2. I have a strong suspicion that if we can figure out and fix whatever kwin is doing strangely with that game, all of the other results will get a nice uplift.

And a personal anecdote, with dual monitor on Xorg, I pretty much have to disable compositing (alt+shift+f12), otherwise I get huge frame drops or low FPS. Not sure why this is the case, but I wonder if these graphs are indicating this issue I experience too.

[V1tol]

I wonder how will kwinft and kwin-lowlatency look in these tests. I think problems with Hitman were because it failed to request compositor to disable for some reasons.

[CochainComplex]

I find it interesting how my experience with Wayland in games is just the opposite of the Phoronix benchmarks, and it's not just about FPS.

Nvidia ?

[bug77]

With active window compositing. There is ~0 rendering performance difference between Xorg with window manager compositing off and Wayland with direct scanout (which currently works more often with Mutter than with KWin, but this should about to change soon) on dGPUs.
If Michael configured his benchmark environment so that KWin compositing would not be automatically enabled on log-in, the Xorg results likely would look different.

These are games, I would expect compositing to go away the minute I switch to full screen. Or does Michael test them in windowed mode?

[Mario Junior]

sounds like you have a lot of things to say

I already said many times here. Stutter, less FPS than Xorg, vsync enabled everywhere, mouse cursor becomes a shit on games...

This happens in whatever distro I use.

Nvidia ?

AMD.

[RejectModernity]

If only Michael could test games with native wayland support, like dota 2 on opengl. Just add SDL_VIDEODRIVER=wayland at the launch options.