Ubuntu 19.04 Radeon Linux Gaming Performance: Popular Desktops Benchmarked, Wayland vs. X.Org

[Brisse]

And the mere fps values also don't show if there is stutter introduced by the compositor or not.

This! It's gotten a lot better after x.org 1.20.4 and with latest GNOME but when running xwayland there are issues with frame presentation that don't show up in benchmarks. At least with older x.org releases, rendered frames that showed up on frame-rate counters were clearly discarded by the compositor before ever reaching the screen because you could have a hundred fps and it still looked like a slide-show. Also the presentation timing could be wrong sometimes, which caused serious headache inducing jitter.

It's gotten much better with the latest x.org, gnome-shell and mutter, but experience from past versions has showed that benchmarks of games through xwayland needs to be taken with a grain of salt because they have historically not represented the subjective experience. The Phoronix Test Suite is clearly measuring the amount of frames being rendered, not the actual amount that is displayed on screen. There can be a big difference there depending on what the compositor is doing.

[eidolon]

XFWM4 has similar functionality [disabling compositing for fullscreen windows], but the bug so far is that the heuristics to detect when something is fullscreen is not triggering for the cases we looked at (the various PTS gaming/graphics tests).

That was the case in July 2016, as noted by Clear Linux developers. Is xfwm4 still affected?

[GrayShade]

Michael, On the next test, can you include sway as a lightweight Wayland compositor?

[Ardje]

I am a fan of Xorg myself, but I don't get the wayland must die attitude. I probably used X11 before some of you were born, and I didn't get the hate some had against X11.
But saying that wayland is not the future for local graphics is sounding like you have a very high end machine and you expect everybody else to have a high end machine too. I am sorry, but I work with 10's of thousands low end machines, and if I can get rid of X11 overhead on those, yes please.
What should be benched is software like chrome (a real X11 overhead creator), or any other application doing lot's of text and graphics mixed.
For now I use wayland because the intel as well as the nvidia drivers are much smoother for programms like mixxx. Drawing waveforms without stutter is a special art it seems on X11.
On armhf I do not have a working wayland + gles environment so it is done in x11 and software at 15fps instead of a steady 60. It really hurts. It feels like I am looking at serato at a midrange computer.

[trek]

one Openbox to rule them all

[finalzone]

There is no wayland gaming videos in YouTube, so it must be really bad in real life. This and many other Phoronix articles distort the truth. Ban for those who disagree the official Phoronix view.

Running The Rise of Tomb Raider on Fedora 29 default Gnome Shell Wayland session powered by Sapphire Radeon RX 560 4GB using amdgpu driver iwithout issue.

[gregzeng]

It was good to see that ALL the "flavors" of the Ubuntu family performed similarly. MATE was surprising "slow", and LXQT was the fastest, generally of the Desktop Environments (DE). If tested with the very latest Ubuntu Linux kernels, I expect these results to be even better. Ubuntu releases updates to the Linux kernel every few days at their official web site. We might also assume that these speedy benchmarks apply to the rest of the Ubuntu derivatives: Linux Mint, Feren, etc.

This Phoronix report should be read in conjunction with the other report:

Intel Xeon Cascade Lake Running Even Faster With Clear Linux - Six Linux Operating Systems Benchmarked"

05 Apr 2019

This implies that the whole inner Ubuntu family (MATE excepted) should outperform most other Linux operating systems:. In that report, tested were: Ubuntu 18.04.2 LTS, Ubuntu 19.04 Beta, Fedora 29, CentOS 7, Debian 9.8, and Clear Linux.
Judging from the comments, your readers do not always appreciate your work. "Fanboys" exist in all technological products, including Linux operating systems. On some of your results, the percentage differences seem so small, that random chance might result in the reported differences. I'm assuming that you give us the average of several results, with perhaps the two outlier extremes missing from the reported results?

The other significant finding for me was the performances of Wayland versus X.org. In this early stage of Wayland's development, there is no speed differences between the two. As Wayland develops, it should be the features (caches, 4k, 8k, etc) that shows the "superiority" of the Wayland.

[Ardje]

Wayland needs The Radeon RX Vega 64 graphics card and Core i9 9900K.

https://github.com/giucam/orbital

" Since it uses QtQuick 2 to draw the interface it will make use of OpenGL, so it is advisable to use a decent graphics driver, otherwise the performance will not be good. ¨

I have used Debian Xfce in a Pentium III tablet PC, so X11 does run well in low end machines. Do not use wrappers like Xwayland, they slow down and cause bugs.

Well, the thing is: if you look at Samsung: https://www.phoronix.com/scan.php?pa...-Loves-Wayland that's all about low power CPU.
Don't get me wrong: all the hate on X11 usually is misplaced by people that never took the time to look into the protocol. Running linux on a 386 was more lightweight than for instance windows 3.11. And X11 is a big part of that impact. There are a lot of daemons build for kde and gnome that X11 already provided but was never used (except by netscape navigator for instance). So taking into account that they already did not adhere to the X11 protocol anyway...
And yes, there are a lot of negatives about wayland. For instance, I can kill enlightenment, and it will restart and manage all windows again without closing my apps. On wayland a single bug in the compositor is fatal. But if you have a limited end use (POS), then that's acceptable.