AMD Radeon Linux Gaming Performance At Parity Between KDE Plasma 6.0 X11 vs. Wayland

[sophisticles]

AMD Radeon Linux Gaming Performance At Parity Between KDE Plasma 6.0 X11 vs. Wayland

I would hope and expect this to be the case.

The few benchmarks under Gnome that do show a difference that can be attributed to inherent code differences should eventually also achieve parity.

The reality is that from the end user's point of view there should be no performance regressions at all from using Wayland.

[loganj]

did u test it with the APU too? on a laptop for example?
i had ugly frame pacing with wayland on my laptop APU where game reported 40-50 fps but everything was working like 10 fps.
i know there is a bug report also about this and not just with APU but also with some dgpu from AMD

[Berniyh]

Thanks for the benchmarks, but I doubt this can shut the blissful ignorants who keep parroting bullshit, since they never look at facts to begin with.

You're quite harsh to yourself there.

[n4d47ecya]

I don't understand what's the point in comparing between modern KDE 6 and two year old Gnome 42?

[ezst036]

Anyway they should stop using the Linux kernel since that's also tech from the 90s. Who cares it got a lot of updates since then, right?

How did it receive updates? Has the Linux kernel been declared abandonware, as other things being discussed right here have been declared?

[treba]

Wayland's issue isn't performance directly, it's lack of features (by design). I say directly, because some things like childing yourself to another window (for Wine's virtual desktops for instance) require more overhead due to lacking features that X11 has natively, so the performance will be lower (although it's not perf directly but input lag and latency). This is due to its stupid design and stubborn devs, and yes, it's the protocol's fault.

What overhead do you expect, assuming a architecture around dmabufs sharing? I don't see why e.g. additional IPC would result in input lag in any human-conceivable form (if we define that at say 0.5 ms).

[ezst036]

Phoronix: AMD Radeon Linux Gaming Performance At Parity Between KDE Plasma 6.0 X11 vs. Wayland

With a few weeks having passed since the KDE Plasma 6.0 desktop release along with some point releases for addressing initial fall-out, I've been meaning to run some Plasma 6.0 Linux gaming performance benchmarks. I'll have up some interesting metrics soon using Fedora 40 while for this initial article is a look at the KDE Plasma 6.0 gaming performance between the Wayland and X11 sessions atop KDE Neon. Then similarly are the results for GNOME Shell with its X11 and Wayland sessions.

https://www.phoronix.com/review/kde-plasma-6-amd-gaming

Yes, but. You know this is the only thing people actually see:

Phoronix: X11 vs. Wayland X11 vs. Wayland X11 vs. Wayland X11 vs. Wayland

[Classical]

bspwm outperforms KDE Wayland, river wm and sway wm on Alpine Linux in Xonotic and 5 Steam games. I think Wayland is still slower and less reliable.

[varikonniemi]

What overhead do you expect, assuming a architecture around dmabufs sharing? I don't see why e.g. additional IPC would result in input lag in any human-conceivable form (if we define that at say 0.5 ms).

x11 mess causes CPU overhead, and if you don't have enough fast enough cores to eat it up you will see even severe performance drop.

granted this can usually only be measured in performance on architectures with slow cores, or very few of them (like 1, 2 with HT)

But in power draw it can always be measured, in some cases x11 can increase power consumption by 20%

[andyprough]

From the article:

From there I ran a variety of gaming benchmarks using the kdefault KDE Plasma 6.0 Wayland session for the (X)Wayland gaming performance and repeating the tests as well under the alternative X11 session.

You're comparing an X11 session to a game running under XWayland which is an Xorg server running as a Wayland client - why would you expect to see a benchmarking difference? All you are showing is that Wayland is just as efficient at invoking an Xorg server as running an Xorg server natively. But isn't that to be expected after all these years of development and refinement?