Fresh RadeonSI Mesa Git Gaming Benchmarks On 7 Linux Desktops

[indepe]

Newbie question: if LXDE and Openbox can do without a compositor, why do the other desktops use one?

Considering the Dota 2 Vulkan result, the Talos Principle Vulkan result would have been interesting.

Surprising that for different games, the results / ordering are so different.

[dungeon]

Someone else commented about unredirection, meaning (as far as I know) that the compositor is trying to buffer the game window.

Yeah that might be first guess about unredirection, but it might not be... to me it is just slow whatever reason is Bug could be on either compositor, driver or app sides.

[dungeon]

Newbie question: if LXDE and Openbox can do without a compositor, why do the other desktops use one?

Because "our customers use compositors" someone said 10 years ago and stayed alive, while same these compositors still has slow cases and are all broken.

Only ignorance is sustainable here, i even think how ignorance is the main feature of opensource. Goal is probably to have VR instead of Desktop with every pixel perfect or something like that, until someone does not introduce another goal

[indepe]

Because "our customers use compositors" someone said 10 years ago and stayed alive, while same these compositors still has slow cases and are all broken.

Only ignorance is sustainable here, i even think how ignorance is the main feature of opensource. Goal is probably to have VR with every pixel perfect or something like that, until someone does not introduce another goal

Haha. I do like the "every frame is perfect" goal. I think. However I do not know what kind of problem they are trying to solve in this context. But then, I don't even know if the benchmarks are running in full-screen-mode.

[dungeon]

Haha. I do like the "every frame is perfect" goal. I think. However I do not know what kind of problem they are trying to solve in this context.

It does not solve anything, it just ignore old and introduce new problems as people like experimenting

[AsuMagic]

Xfce is stable, freely configurable and ready. KDE, Gnome3, Unity, Budgie not because they are under heavy construction and barely tested.
Why waste 500MB for a dekstop, XFCE vs KDE,Gnome3 and Budgie. Xfce runs in Pentium 3/ 800Mhz, 500MB RAM computer, KDE, Gnome3, Unity, Budgie not.

I boot my gnome 3 laptop to 200MB RAM on wayland + xwayland and i still could reduce it

[Otus]

Only ignorance is sustainable here, i even think how ignorance is the main feature of opensource. Goal is probably to have VR instead of Desktop with every pixel perfect or something like that, until someone does not introduce another goal

I'd rather take pixel perfect frames and a guarantee of tear-free vsync-locked display in exchange for a 20% or even a 50% performance loss. But then again, I don't play twitch shooters so I realize my requirements are not the same as everyone else's.

I never understand why people complain when other DEs do things differently. That's the whole point. They serve a different niche. Some people want performance; some want eye-candy; others just want the software to keep working as they are used to.

[jones13]

Michael, I have a blocked comment.

Also, can I suggest a new form of benchmark? In a few weeks you will test the new Ryzen, but a lot of people (myself included) are asking themselves if worth to buy a 8 core CPU for gaming.

So, it could be interesting to do a benchmark called "how many cores you need" kind of test. Disabling cores in the BIOS, you can see how many cores is needed to achieve the best performance.

This is a rare type of test to be found in the internet, and with the upcoming Ryzen a "affordable" high performance 8 core CPU is coming, so it would be nice to see if worth take it or wait for the 6 and 4 core ones, for gaming purposes.

The tests can be done now with that 8 core, 4 Ghz Xeon of yours.

Computerbase.de has recently published an article to this topic in preparation for upcoming Ryzen benchmarks.
They benchmarked of course only on windows.

https://www.computerbase.de/2017-02/cpu-skalierung-kerne-spiele-test/#abschnitt_performancerating_in_full_hd

(German language but you can look at the diagrams)

Conclusion is that more cores/threads with lower frequencies become slowly more important than less cores with high frequencies.

As said these are windows only results. For linux game ports the results could be completely different.

[indepe]

I'd rather take pixel perfect frames and a guarantee of tear-free vsync-locked display in exchange for a 20% or even a 50% performance loss. But then again, I don't play twitch shooters so I realize my requirements are not the same as everyone else's.

I never understand why people complain when other DEs do things differently. That's the whole point. They serve a different niche. Some people want performance; some want eye-candy; others just want the software to keep working as they are used to.

It might be just my ignorance, but I'd be surprised if even a 1% performance loss, or any noticeable one at all, except an *occasional* 1-frame-later display in specific situations, is necessary to get a tear-free desktop. Of course, it might require some changes. However it's not like I would know the specifics of why it might be difficult to achieve a 'tear-free' desktop without noticeable performance loss.

[Mike Frett]

XFCE has a fairly severe compositor bug that greatly impacts performance. Given that, I think any new tests should involve disabling the XFCE compositor.