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

[theriddick]

Assuming this is with full Explicit Sync set of patches.
Wine/Wayland/AMD Driver. Full Package has explicit sync now?!

[oiaohm]

Assuming this is with full Explicit Sync set of patches.
Wine/Wayland/AMD Driver. Full Package has explicit sync now?!

The answer will be no because different sections of Wine does require implicit sync. Fun of your old Windows games.

Reality is you cannot do everything as explicit sync. Question is after all the patches will we have as much explicit sync as what is sane to have. The issue with eglstreams is it had more explicit sync than what was sane.

There are different areas in Zink when it doing implicit sync with opengl where it has a performance problem because due to vulkan immutable only option it has it to make a new buffer. Implicit sync with GBM/DMABUF ... Opengl there are transformations you can do on buffers that you have to forbid in explicit sync mode because if you don't you will suffer from instability.. Working around these forbid operations by buffer duplication as zink does has a higher processing cost.

The fun of running old opengl and direct x legacy games.

Nvidia saying we will not support implicit sync is saying we will not support legacy games and programs.

The reality here is the stack will need a mix of implicit and explicit sync. We want it bias in the explicit sync direction but we have to accept implicit sync for particular is required and is just not optional for performance and stability.

[oleid]

Show me where it was declared officially, not by a Wayland clown.

What do you require? An official statement by the Xorg foundation?

I guess you are aware of the developer comments the OP was referring to.

https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/533#note_670522

https://twitter.com/danvet/status/1320708172550688769?s=20

And there were articles in several news outlets on that matter.

https://www.phoronix.com/news/XServer-Abandonware

I guess it is just a matter of reading between the lines in lieu of an official statement.

[treba]

X11 has IPC too, but Wayland would have more IPC, because now the process that nests child windows of other processes has to do composition itself, and then IPC back to Wayland.

In X11 there's only one IPC, from all processes to the X server/compositor.

In Wayland processes first have to IPC to the process embedding them, then that process IPCs to Wayland compositor. Because it does not allow child windows to be from another process. Artificial limitation btw, probably has even more code to block it!!

Yes, in the fucking protocol. Also, see, apps (and malware) can workaround it, so what the fuck is the point of it?

I was asking why you think that would be a problem? I'm pretty sure input IPC is pretty cheap and fast. Browsers do it for website isolation as well. What use-case for child windows do you have in mind that has higher latency requirements than webgl games?

As for the question of why the Wayland protocol does not allow child windows: I guess in order to keep server complexity low. If stuff like that can live in some client library, then it should.

[mSparks]

I see, so your anger and negativity are because of your own failures in writing better code. You are left gutted huh? ... whatever it means

In my opinion, Wayland is the best thing happened to Linux in a long time.

How can anyone be angry with people showing off the next linux display server, almost but not quite as good as current linux display server.

gutted
adjective

INFORMAL•BRITISH

bitterly disappointed or upset.

"I know how gutted the players must feel"

[MrCooper]

As far as I know there shouldn't be major differences between Xorg and Wayland performance.

Indeed, there's no inherent reason why there would be any significant difference for this kind of benchmarks. The GPU drivers and apps themselves should affect the results a lot more than the display protocols, which work mostly the same for direct rendering.

Also it seems like the drops in performance GNOME has are only under specific games. They should profile that but I suspect its just Mutter still not being efficient enough.

If something in mutter caused it, why didn't it affect all tests consistently? E.g. CS2 took a 20% hit at 4K, none at all at 1080p though (GNOME Wayland even took the overall win there).

Also, some of the tests required N=12 runs to reach "reasonable" standard deviation.

This is evidence pointing to some unknown factor affecting some of the test runs.

Assuming this is with full Explicit Sync set of patches.

That would require a time machine.

[varikonniemi]

I know it can seem threatening for a protocol to be 15 years in the making before being extended enough for all desktop use cases. But this happens only because it does things right and not fall into the same shit category as windows and macos.

[piotrj3]

The answer will be no because different sections of Wine does require implicit sync. Fun of your old Windows games.

Reality is you cannot do everything as explicit sync. Question is after all the patches will we have as much explicit sync as what is sane to have. The issue with eglstreams is it had more explicit sync than what was sane.

There are different areas in Zink when it doing implicit sync with opengl where it has a performance problem because due to vulkan immutable only option it has it to make a new buffer. Implicit sync with GBM/DMABUF ... Opengl there are transformations you can do on buffers that you have to forbid in explicit sync mode because if you don't you will suffer from instability.. Working around these forbid operations by buffer duplication as zink does has a higher processing cost.

The fun of running old opengl and direct x legacy games.

Nvidia saying we will not support implicit sync is saying we will not support legacy games and programs.

The reality here is the stack will need a mix of implicit and explicit sync. We want it bias in the explicit sync direction but we have to accept implicit sync for particular is required and is just not optional for performance and stability.

Nvidia doesn't have problem supporting at all OpenGL or old Directx games. In fact stability wise they exceed here AMD on average. Stop saying FUD. The only problem Nvidia has is compositor level calls.

[oiaohm]

Nvidia doesn't have problem supporting at all OpenGL or old Directx games. In fact stability wise they exceed here AMD on average. Stop saying FUD. The only problem Nvidia has is compositor level calls.

Its not fud. The issues Nvidia has had with glamor due to the way glamor are the same that happens with some your old opengl 1.0-.2.0 games. These are design to exploit the locking of implicit sync.

https://gitlab.freedesktop.org/xorg/xserver/-/issues/1317

Yes early direct x games as in direct x 1.0 that wine can run end up using opengl or X11 2d rendering while expecting correct implicit sync. When implicit sync is not right as that issue 1317 describes everything turns into a mess.

Yes there is a reason to use zink on Nvidia hardware to get a better opengl implicit sync implementation.

[mrg666]

While It's nice to see XWayland (X11 + Wayland benefits) currently being on parity with X11, I think we still need some key software to migrate to Wayland before making a objective "X11 vs Wayland" comparison. Those being:

• Steam (CEF)
• Proton (Wine)
• Native games (mostly SDL)

That aside, I'm very happy for KDE users about the success of the newest version.

Steam and proton are coming to Wayland; they are working on it. But I doubt many AAA games will be native. Why would they do it while wine/proton will run that game on Linux with good performance?