X.Org vs. Wayland Browser Performance With Firefox + Chrome

[Volta]

kozman yes, Google has a lot of resources to spend on spying us. Does it make it better? I doubt.

[Guest]

GNOME is the default/standard desktop!

What!!!
Don't drink before turning on your pc .

[bug77]

Multiple things:

At first Wayland isn't just about performance, it's also about security, maintainability and other things. Performance is just one of many aspects. So even if it would fail to deliver on that promise the others might still work.

Second, synthetic browser benchmarks are still crap. They measure some specific workloads that aren't too common to happen when surfing the web. I dislike most all of them, without putting chrome/firefox or X11/wayland/windows, etc into my equation. It's more about real user experience, so compare yourself.

Just one example: In Wayland we'll finally get video acceleration for standard video streams. You'll never see this in the graphs of those benchmarks, but you'll perceive a noticeable benefit.

Third, from michaels sum up:

"The WebXPRT results are quite promising for the Linux desktop on Wayland with better performance compared to the traditional X.Org sessions. The other benchmarks meanwhile generally showed Wayland at least comparable to X.Org and does point towards Firefox having a stronger footing on Wayland at this stage compared to Google Chrome."

After reading the graphs and reading that, I suspect Wayland to deliver on its promise. Some Wayland features and tweaks are also still hidden behind about:config flags in firefox and not activated by default. I suspect Michael tested without using all bells and whistles, and still got good results, so you can expect to get even more performance in the future.




MOZ_ENABLE_WAYLAND="1" firefox

One thing: a decade after its inception, it still does nothing for performance. So far it only fixes security issues never documented to cause something significant in the wild (still welcome, as is any security fix) and does away with being able to run the same across GPUs (not Wayland's fault, but that's where it got us).

Fwiw, systemd is 2 years younger and has already achieved a lot more.

[Vistaus]

Add Ubuntu, Suse, and Arch/Manjaro and you have 99% of the installed systems running a GUI.

Ubuntu? Arch? Manjaro? Nice try, but Ubuntu doesn't use Wayland by default, same goes for Manjaro and, due to its nature, Arch lets the user choose whether he/she wants Wayland or not.

[Vistaus]

So did these really run on Wayland and not Xwayland? You still have to set variables for Firefox on Wayland afaik

No, they ran on XWayland 'cause Michael is not an expert and doesn't know what he's doing... [/sarcasm]

Come on dude, have a little faith in Michael I'm 100% convinced that this was a fair test.

[Britoid]

Ubuntu? Arch? Manjaro? Nice try, but Ubuntu doesn't use Wayland by default, same goes for Manjaro and, due to its nature, Arch lets the user choose whether he/she wants Wayland or not.

Arch rightfully uses upstream defaults, which means Wayland on the GNOME package.

[duby229]

Do people forget that Wayland wasn't about huge performance gains? Like that wasn't why they designed it. It was designed to be a protocol that was designed for the modern computing era that took into account the way the software landscape was, like GUI toolkits and no one using the X primitives, and also the fact that lots of the core things, like input, just needed to be completely rebuilt which couldn't be done lest you break all the old stuff. If you see a massive performance gain by using Wayland its going to be something they couldn't do in Xorg, like the proper GPU accel stuff

Ha! Considering your argument about core things like input, and input latency on Wayland is almost completely unbearable.... The bottom line is core things like input is exactly what xorg did right and Wayland totally fails at.

[Volta]

Ha! Considering your argument about core things like input, and input latency on Wayland is almost completely unbearable.... The bottom line is core things like input is exactly what xorg did right and Wayland totally fails at.

Are you sure it's Wayland problem or perhaps Mutter? Comparing X to Wayland is like comparing apples to oranges. Wayland isn't X server drop in replacement. It's a different approach and there are changes needed on many levels. Some people seems to be very confused.

[arokh]

Very pleased with Firefox 76 beta performance on Sway here, thanks to everyone who contributed code and put in the work so we can have this stuff for free. The few but very vocal anti-wayland lunatics are becoming just noise at this point, hopefully they will join Devuan or BSD together with the rest

[duby229]

Are you sure it's Wayland problem or perhaps Mutter? Comparing X to Wayland is like comparing apples to oranges. Wayland isn't X server drop in replacement. It's a different approach and there are changes needed on many levels. Some people seems to be very confused.

I'm positive for sure. Most of the major distro's seem to have mitigated input lag somewhat by implementing (cgroup) rules to give Wayland maximum CPU under load conditions. But that's not a fix. That doesn't fix the input lag. It just brute forces past it.