Ubuntu 20.04 Gaming Performance Across Desktops, X.Org vs. Wayland

[Volta]

Xorg is alive and kiching, the brain of those wayland's fans is dead.

This is false information, but in your delusional world it may be true. What Xorg Project Team is saying about XWayland:

Support for legacy X11 clients running inside a Wayland session

X11 is legacy and this is confirmed, so..?

[Volta]

but X (and all the components of it) are definitely still being maintained, so knock off the FUD.

Yes, it is in maintenance mode. That's the whole point here, so thanks for confirmation it's not being developed anymore. There's a lot of FUD in this thread regarding Wayland, but it seems not everyone is able to face the reality.

[Guest]

This is false information, but in your delusional world it may be true. What Xorg Project Team is saying about XWayland:



X11 is legacy and this is confirmed, so..?

Ok, where? Or it is in your mind only.

[Volta]

Ok, where? Or it is in your mind only.

Rent a secretary if simple tasks are too much for you.

[Britoid]

Ok, where? Or it is in your mind only.

Are you dumb or just trolling?

The people behind X are now working on Wayland and have declared no future major work is going into X, with Wayland being the current thing.

[Britoid]

All I've been hearing from Wayland adepts for the past twelve years is "soon". Meanwhile 12 years in we have a single working DE for Wayland which is Gnome. Let's be patient for 10 more years, right? I guess by then it'll be ready to replace Xorg.

You do not design a new graphics subsystem to replace the old "broken" one without all the necessary features baked in right from the beginning.

You do not design a new graphics subsystem which requires a whole more effort and code to make it compatible with your DM/DE (as indicated by Wayland uptake so far which has been horrible, i.e. missing).

You do not design a new graphics subsystem which changes all the conventions so much, old applications need to be substantially altered to work efficiently/fast under it.

You do design a new graphics subsystem which offers obvious benefits where the one before it struggled, i.e. under Xorg both the X11 protocol and VNC are super inefficient, yet Wayland is just the same crap squared: everything on the screen is rasterized and you cannot efficiently send over the pipe. Sorry, that's just crap.

FYI Windows has updated their protocols many times, but since everyone on Windows uses the Win32 APIs for window creation etc, they get updated too. Unfortunately we don't have that luxury on Linux because outdated toolkits are talking directly to the X server.

[allquixotic]

All I've been hearing from Wayland adepts for the past twelve years is "soon". Meanwhile 12 years in we have a single working DE for Wayland which is Gnome. Let's be patient for 10 more years, right? I guess by then it'll be ready to replace Xorg.

You do not design a new graphics subsystem to replace the old "broken" one without all the necessary features baked in right from the beginning.

You do not design a new graphics subsystem which requires a whole more effort and code to make it compatible with your DM/DE (as indicated by Wayland uptake so far which has been horrible, i.e. missing).

You do not design a new graphics subsystem which changes all the conventions so much, old applications need to be substantially altered to work efficiently/fast under it.

You do design a new graphics subsystem which offers obvious benefits where the one before it struggled, i.e. under Xorg both the X11 protocol and VNC are super inefficient, yet Wayland is just the same crap squared: everything on the screen is rasterized and you cannot efficiently send over the pipe. Sorry, that's just crap.

I don't think Wayland is fundamentally designed badly. What it is, is incomplete.

My two central theses are: (1) To overhaul a display system will require more than a decade, and (2) backwards compatibility is paramount.

XWayland will be absolutely critical to the success of Wayland, because proprietary applications exist on Linux (pro tools like CAD; games; legacy line of business apps), and because even open source projects take significant time to adapt to a new display system. Wayland will be ready for prime time when XWayland is feature-equivalent (and absent any breaking bugs) with native Xorg -- but not before.

This struggle is not unique to the Linux desktop. We're trying to solve a hard problem that others have gone through before.

The old Windows XP graphics system took about 10-12 years for Microsoft to replace, and that's with a lot more focused/concentrated development. They got it so wrong that -- if I remember my history correctly -- the graphics stack is a major reason why the "Longhorn" pre-release of Vista had to be scrapped and rewritten mid-development.

Then, Vista's production release ended up being a lot more like XP than they wanted, and it basically took until Windows 10 to eliminate the pre-WDDM stack. And they still go through an unthinkable amount of contortions to maintain ABI and API compatibility with old Windows binaries that use GDI+ and such, which is their equivalent to XWayland on Linux.

So to replace XP's graphics stack with a new composited graphics stack ended up taking well over a decade for a company whose livelihood has been built on having a stable GUI-driven desktop.

In my opinion, the consistent push from Red Hat and Canonical, and eventual Nvidia driver support for accelerated windows in XWayland will result in an inevitable transition to Wayland. But by no means should your average Linux user just stop running Xorg and switch to Wayland today. I would guess we are still at least 1-2 LTS releases of Ubuntu (so, 2-4 years) away from Wayland as default, with Nvidia driver support being the biggest determinant. BUT, that doesn't mean Xorg is going away.

I would conservatively say that the Xorg display server has about 10 years left of wide adoption. I would say that the X11 protocol as it is used in applications has at least 30 years of life left in it, meaning it will be totally normal 20 years from now to still run a lot of X11 processes in XWayland. But by then, XWayland should be so smooth that you won't even notice you're running something on XWayland vs. native Wayland. Just like you don't notice when a Windows app is running on GDI+ vs. the newer stuff (Direct2D/Direct3D, etc.)

[treba]

Interesting what people read into this. What I take from this is: the benchmarks help uncover bugs - in cases where things work fine the differences are barely measurable any more.

[birdie]

FYI Windows has updated their protocols many times, but since everyone on Windows uses the Win32 APIs for window creation etc, they get updated too. Unfortunately we don't have that luxury on Linux because outdated toolkits are talking directly to the X server.

Classic win32 applications worked blazingly fast under Windows NT 3.1 (released in 1993) and they continue to run that way under Windows 10 64 in 2020, i.e. 27 years later. Oh and don't get me started on RDP which was created around Win32 APIs and works a little bit like libX11 only way faster and more efficiently.

On the other hand Wayland was designed solely to push bitmaps to the screen - this sounds almost like a bad joke to me. What a commendable initiative! Only that's not how Windows, MacOS X, iOS or Android work - all well established OSes with a proven track record. To be honest Android on the surface looks like Wayland only it has system wide drawing APIs which Wayland lacks in any shape or form. Skia could probably become its alternative for Linux but Skia alone is not sufficient (you need to render fonts as well). And Skia still pushes bitmaps and we're back to square one: a bitmap pushing protocol.

[birdie]

Interesting what people read into this. What I take from this is: the benchmarks help uncover bugs - in cases where things work fine the differences are barely measurable any more.

No one is arguing about that. What people are arguing about is that such bugs still exist in 2020. And they are not the only ones. People often report rendering issues for Firefox under Wayland and other niceties. You'd think in 12 years they could have ironed that all out but here we are.