GNOME X.Org vs. Wayland Performance + Power Usage On Fedora 32 With AMD Renoir Laptop

[paupav]

wayland is so much smoother, maybe thats the reason. can it be made to be as choppy as x and have tests then.

[blacknova]

It works, but it does not work well and there is no way to fix it.
The only way this could actually be solved is to write a completely new implementation of the X11 Specification, but then we can just stick with Wayland, it already has 2 good implementations.

It does work well enough for now at least. Yes, in long run we all will switch to wayland probably since were are no alternatives around or in planning. But right now, it still have enough issues to now do that at least for workstations. It might be quite all right for home users though, well, unless you have nvidia. X11 specification is out of date by decades already and while actual idea is good one to described functionality is not up to the current tasks, after all it is why Wayland was born. The problem with Wayland is that instead of updated functionality to more recent and deciding on security measures, it was decided to completely ignore benefits of server side rendering and push all rendering to client, security was updated though.

On multiple implementations - I'm not happy about it at all, how long it would be until these implementations become incompatible enough that software writers would need to take these differences into account?

[finalzone]

Wayland doesn't work today, X11 does. Both are crap but one works now. Hopefully Arcan can fix this mess since Mir is dead.

Maybe that custom Ubuntu version implemented poorly Wayland as Mir got mentioned.
The reality is Wayland protocol is already in use in production from enterprise (i.e. Red Hat Enterprise Linux 8 Workstation series) to embedded like Tizen (on Samsung Galaxy Watch).
This message is written on Fedora 32 running Fedora Firefox Wayland via GNOME Wayland.

Name Firefox
Version 77.0.1
Build ID 20200603192437
Distribution ID fedora
User Agent Mozilla/5.0 (X11; Fedora; Linux x86_64; rv:77.0) Gecko/20100101 Firefox/77.0
Window Protocol wayland/drm
Desktop Environment gnome

[reba]

To check if you run firefox in wayland mode, check about:support for "Window Protocol"
There are two options, wayland and wayland/drm depending on your about:config settings.
Testing in my Pinebook Pro with Gnome 3.36, so actual lowest end hardware.
Xorg/Xwayland is unusable, completely unusable.
"wayland" is a big improvement but not smooth.
"wayland/drm" is damn smooth, it is running near to perfection and scrolling just makes fun.

It could be that Michael tested accidentally in Xwayland mode, or maybe he should try again with wayland/drm mode.

I probably just have not seen the test environment and its settings but
Michael, maybe we could get a clear answer to this question, which is essential to get a feeling for the reported Firefox performance you've shown us.
Please don't add needlessy heat by omitting this, I am positive it's also in your interest to keep the forums calm.

[tuxd3v]

So canonical was right about wayland..
wayland is pure garbage, compared with xorg..

Bring back mir, with its native protocol please.. because wayland consumes more power, more memory, is slower, it has nothing good in its favor..

Some said X needs to be replaced because its big and resource intensive?.. what???
wayland is far worst..

mir was, after all, in the right track!

[smitty3268]

Michael - can you explain the results of the 4th and 5th Firefox tests in the article?

In the first set of (5) charts that display the speed, the results seem pretty close on the 4th and 5th tests.

But in the 2nd and 3rd set of charts that display power and memory use, it seems to display the Wayland session tests in the 4th/5th charts ending much faster than X11?

Were these different tests? Separate runs of the same test? Are the results that variable across runs?

[raster]

I have to say .. it's rather amusing to see this debate... and from my guesses primarily from people not involved in either X11 or Wayland development... and such extreme opinions from people who aren't actually working on it and all focused on a single Wayland compositor implementation only...

Some facts: Wayland probably drives more displays than X11 does at this point, and even if it doesn't, its growing substantially and quickly and is already substantial. Every Samsung smart watch, smart TV and fridge run a Wayland compositor and has done so for a while. That'll be 100's of millions of devices by now. Wayland is real and in production. I do not know numbers from the car infotainment world, so can't say.

For people saying "Why don't you just design something more like OSX or Android", well, Android uses surfaceflinger as its compositor and display system and surfaceflinger is incredibly similar to Wayland in design. They are almost twins. At least in comparison to X11. But let me get to the fun bit.

This is NOT an exhaustive test. It's very simple and I spent only 5 mins getting the numbers (spent much more writing this comment vs getting data), BUT it's a different compositor to Gnome, but like Gnome the experience is "similar" in that it's the same DE/WM/Compositor with same config, same feature-set on the same piece of hardware with the same OS - the only difference is I logged into a text TTY and then either ran startx or just the WM/Compositor directly, then ran 1 terminal and checked for free available memory. About as apples to apples as it'll get, much like Gnome. So not an exhaustive set of apps and so on, but let this be a different point to "I sampled only 1 environment". No - I'm not going to spend hours formulating some battery life tests and others and run it etc. I'm only going to go this far as I have other things to do like sleep. This is just a "there's another side to this coin and take a look here to get an idea".

This was Arch Linux (x86_64), recently updated in the last week or so on an Intel i5-4200U CPU (and GPU) with 4GB RAM, with current git master of EFL+Enlightenment+Terminology (because it's what I work on). Same theme, same user, same config files, resolution, screen etc.

First. We need to find the base system memory usage to log into the TTY and this is a base memory footprint that is shared between both and should be subtracted as it is not relevant to the different display systems, so the simple test, running "free -m" to get system-wide footprint:

total used free shared buff/cache available
Mem: 3844 90 3607 1 146 3562
Swap: 2047 0 2047

So 90M etc. as a base. Now Let's look at X11 (startx):

total used free shared buff/cache available
Mem: 3844 250 3169 62 424 3312
Swap: 2047 0 2047

250M for X11. Now let's look at Wayland:

total used free shared buff/cache available
Mem: 3844 188 3278 36 377 3400
Swap: 2047 0 2047

188M. So X11 vs Wayland:

RAM SHR BUFF
X11: 160M 61M 278M
WL: 98M 35M 131M
WL/X11: 61% 57% 47%

So... about 40% or BETTER memory savings for WL vs X11. Please calm down a bit and actually get stuck into details and research. If I can, in 5 minutes, show the exact opposite results with a close-as-possible apples vs. apples comparison, then I think you should allow room for there to be vastly differing results depending on what you choose and that Wayland is actually capable of being much better than X11. I write the above because I work on this code so I know what I'm doing. I still use X11 every day because Wayland support is "not there yet", but the direction of flow is clear. I know where it's at and what it does and what is still needed. I know my comparison above is fair (though very limited) because I know the code behind it on both sides. I know the Wayland support could be trimmed down a bit (it also still runs Xwayland in the background always even if not needed for example, and was/is not needed here) so numbers could be even better. Compositor and clients both share the same toolkit here because that's the design. As someone who works on this, Wayland is still pretty new and just beginning to be adopted and worked on really in the scheme of things. Getting a whole ecosystem to change direction is hard - it's worse than trying to change the direction of the titanic. It's trying to change the direction of a whole flotilla of various boats from titanic to dinghy in size, each with their own captains, crew and hecklers on board. It happens slowly because no single entity is in charge and says "We go this way, like it or not".

As an aside, I do think there is something to be said about a "Widget server" style design which can bring even better results if done well/right but don't give Wayland all the beating without doing homework. Please do it. You may find that differences have more to do with the choice of your DE than Wayland itself at all.

You may find that Gnome Shell on the same system will use many times the memory footprint as well so it's a different implementation and design and thus will net you different results. I think the Gnome guys are doing a great job of pushing Wayland forward faster than most others, so give them credit where credit is due. This will all take time, but let Enlightenment be the shining light of "If you spend your time on the nuts and bolts and get it right, there are some serious wins to be had - look here".

[Volta]

Volta - "can you explain why 30 years later I still can't use alt+Tab with Quake live in Steam?" Because you don't know how to play or configure Quake?

There's nothing to configure. It plays nicely under Wayland.

[jrch2k8]

This is meaningless. Launching Xwayland on demand may save a few mb of ram, but it does not influence Wayland Clients at all.
All you need to care about is if Firefox is running in Xorg, wayland or wayland/drm mode.

Yes, you are partially right.

If you have a 100% wayland only compositor that only spawns XWayland when is necessary the fact XWayland is up won't affect pure wayland clients BUT Gnome isn't there yet and i'm not sure when and why offload things to XWayland.

For example i noticed with certain versions of wayland firefox/Gnome i get sudden CPU spikes on the Xwayland process even when i made sure nothing else is running that could trigger XWayland(aka no X11 apps).

[acobar]

In lieu of change, devoutly religious people will defend the way it was regardless of any evidence presented and drastic worldly changes can be required for devoutly religious people to accept change; like Catholics and the flat Earth and Earth being the center of the universe and how we're STILL working on that.... It's been over 400 years and we're still trying to "prove" a spherical Earth.

X11 supporters can be seen in the same light in regards to moving over to Wayland or accepting Wayland as the way forward. It took them all those years to get their religion where is is now and someone wants to come in and replace that with something new??? Yeah, the Roman's never really made it past Hadrian's Wall so some still believed in the old gods and held out, but eventually, a millennium later, Rome's priests did make their way into Scotland and Ireland and the last of the holdouts were converted into believers of the new god.

We still celebrate Christmas in December because of pegan ritual holdouts. Our days of the week, except one, are all named after Norse gods. The one day is Saturn. Saturn's Day. Saturday. So even 2000 years after the Romans invaded England and started pushing Christianity on everyone, our days of the week are still holdouts from the old gods. Just some more examples of how religious people are resistant of change and that sometimes an XWinterSolstice is necessary so you bring in enough of the old gods to get people STFU about it all. Are you still getting high and boning in the dead of Winter? Yes. Then STFU. XWinterSolstice is code for Xmas for those who can't keep up.

That said, a lot of us are fully aware that Wayland isn't 100% ready for every use-case and that it also took X11 many, many, many years to get where it is to cover all these varied use-cases. The people expecting Wayland to get to where X11 is overnight has always been, to me at least, a laughable stance. These days, though, Wayland is ready for most use-cases so people against it have to dig for very specific bugs and examples to point out why Wayland is wrong or bad or not the way forward. We could all scour GNOME & KDE bug trackers and point out random X11 bugs and spew the same fud to push an anti-X11 agenda.

That's another way they're alike -- they both tend to spew fud to push their agenda.

Anyways, that's the joke/explanation. It helps to be able to see things a bit abstract to get comments like that and this.

"Flat Earth" is something that no educated human being believe since Greek times, Eratosthenes of Cyrene calculated Earth circumference some ~2350 years ago, was really good for that time, and it was seen as spherical even before that, at around 500 BC. Somehow, people keep repeating this lie.
https://www.aps.org/publications/aps...06/history.cfm

I have seen here both sides attacking each other, and even though it is sure that X has been showing its age and got patched to run acceptably on modern hardware, and displaying all its deficiencies, it is also clear by now that Wayland went too far for simplicity, assuming that any lacking trait could be fixed later, and now we have this weird situation where, besides events, to it only exist a video memory buffer. As a consequence, the Wayland protocol is very inefficient to send a screen over network, and we may end up using VNC instead of something like Windows RDP. Worst, every Wayland client must "reinvent" the wheel, what is wasteful and make it hard to have "optimal" results. Will this be addressed? Really don't know, as we have too many groups very invested on their own solutions.

About week days names, perhaps, what you said may be true for English language, but it is not for the others, for Latin derived ones, the origin comes from gods and celestial bodies. See https://en.wikipedia.org/wiki/Names_...ys_of_the_week.

Really, I understand that you tried to use funny parallels on your comment, but I see no similarities.

I'm in no way offended, I'm agnostic.