Running NVIDIA On GNOME's X.Org Session May Get A Lot Smoother

[Britoid]

I see you have never defended a brand. If you see GNOME developers using Apple computers, those developers are sending the next message: "Look that, they don't use their own product, so their product has to be as bad that it isn't worth it to use in real environments."

This is not question about free or non-free software, it's a commercial issue, and the reality of the commercial situation is what I exposed in the previous paragraph.

We actually have yet to see any proof of this, so it's just another one of that guys lies.

But hey, anythings okay as long as it brings in ad revenue.

[DeeZiD]

I totally agree with that. I tried Gnome 3.32 in Fedora on my 9700K+GTX1070 box. And it turned out that it is really not smooth and input lag is quite noticeable under load.

Wait, isn't that using Wayland + Nouveau?
That driver is slow like hell on my 1080Ti even for simple 2D stuff. Also Fedora doesn't have any VanVugt patches yet.
To try out these you either have to compile mutter yourself with like 20 or even more commits he made or just use Arch (using Manjaro here) with patched mutter from AUR or Ubuntu/PopOS with the mutter PPA. Gnome 3.32 without these patches is plain unusable to me, even worse than XFCE or KDE.

With these patches it basically matches Windows 10 smoothness and input lag even while rendering from Resolve 16 which usually results in 100% CPU and like 80-90% GPU load.

[Britoid]

Also Fedora doesn't have any VanVugt patches yet.

Why would it? They're untested, unmerged, unreviewed. Fedora prefers things to be fixed upstream, rather than patching it downstream where further issues can arise, and someone isn't sure whether to file an issue with GNOME or the distro. Canonicals patches usually get found to have severe regressions.

[AsuMagic]

I totally agree with that. I tried Gnome 3.32 in Fedora on my 9700K+GTX1070 box. And it turned out that it is really not smooth and input lag is quite noticeable under load.

And it is really smooth, even under heavy load. That's why I have Hackintosh on the same box as my primary development environment. Yes, in Postgres or WhateverDB performance tests it loses to any Linux, and it is not good in OpenGL, but it is a perfect environment for a web or mobile development. Even Hackintosh is rock stable and capable of running compilation in Docker images, playing 4K Youtube video, debugging in VSCode and chatting in Slack at the same time - without loosing responsiveness at all.

So can I on a perfectly standard kde desktop?

[torturedutopian]

Well, that's great. I was skeptical at first, as a mostly KDE guy, but now we can see Canonical making a really significant difference. A few releases back, Gnome Shell was, in terms of performance, IMHO, barely usable, which actually surprised me a lot -- it was as if no-one was actually using it ?! After quite a few releases of successive improvements & optimizations, it's starting to really pay off (trying 19.04 and it's clearly way better than 1 year back and it seems it will improve A LOT with the recent patches). I'm quite grateful as Ubuntu still is the synonym of Linux desktop for the casual / mainstream users. And nowadays, even joe users are used to a super smooth desktop experience !

Of course, I'm still a bit bitter because the whole Gnome 2 -> Unity -> Gnome 3 transitions sounds like a lot of wasted time. What counts is it's getting really good now ! (as the KDE 3 -> 4 -> 5 transition)

[L_A_G]

Is this true?

It's actually based on a single presentation given by a GNOME developer on a Macbook running MacOS.

People then extrapolated that this was not only representative of what said developer used as his dev machine, but what all GNOME developers used.

So in other words it's really just a meme, sort of like the supposed story about how NASA spent millions developing a pen that worked in space while the Soviets just used a pencil. In reality the famous "Space Pen" was developed by a pen company completely on their own, then offered it to NASA for free as a replacement for the pencils they were using. Even the Soviets replaced their pencils with ones like it after they bought one and copied it.

[DeeZiD]

Why would it? They're untested, unmerged, unreviewed. Fedora prefers things to be fixed upstream, rather than patching it downstream where further issues can arise, and someone isn't sure whether to file an issue with GNOME or the distro. Canonicals patches usually get found to have severe regressions.

In case you didn't notice- this thread is about the latest patches for mutter, which make Gnome faster and way more responsive than other Linux based desktops including stock Gnome 3.32. These commits are very likely to land in Gnome 3.34, in fact many are merged already.

So can I on a perfectly standard kde desktop?

Honestly, even KDE 5.16 has more input-lag on my machines (Intel HD630 or NVIDIA 1080TI or 980GTX) without any load on the CPU and GPU than Gnome 3.32 with VanVugt patches does when both are maxed out. Not talking about framedrops though.

Still didn't try out Kwin lowlatency though...

[Danielsan]

I saw a lot of people making presentation of Linux and alsoLinux DE, with a Macbook and MacOS... Then you understand those people don't really care about GNU, Linux and Floss in general. They are just doing a job. While they can do whatever they want I don't trust people that while are explaining how good is a Gnu/Linux desktop are using a Mac (or Win).

[Britoid]

In case you didn't notice- this thread is about the latest patches for mutter, which make Gnome faster and way more responsive than other Linux based desktops including stock Gnome 3.32. These commits are very likely to land in Gnome 3.34, in fact many are merged already.

3.34 hasn't been released yet. I'm referring to back-porting untested (beyond a few minutes) unstable patches, which Fedora will never do.

I do hope any un-merged land in 3.34, as long as they don't cause problems. Another big thing I hope to land is a GJS garbage collector fix.

[mmstick]

3.34 hasn't been released yet. I'm referring to back-porting untested (beyond a few minutes) unstable patches, which Fedora will never do.

I do hope any un-merged land in 3.34, as long as they don't cause problems. Another big thing I hope to land is a GJS garbage collector fix.

Daniel van Vugt has submitted all of his patches upstream. These patches have not been backported to the 3.32 packages in Ubuntu. I've been backporting and testing these patches on a wide range of hardware at System76 for Pop!_OS, however. We will land these patches early if it can improve the desktop experience for our customers and users today.