Optimize Linux Gaming Performance? | Windows vs Linux

► Follow me!

----------------------------------------------------------------------------------

Links:

(* All Amazon Links are affiliated. That means that I get a small cut of revenue if you buy something, without any additional costs.)

----------------------------------------------------------------------------------

🎵 Music from StreamBeats by Harris Heller.

----------------------------------------------------------------------------------

Chapters:

00:00 - Optimize Linux Gaming Performance | Windows vs Linux
01:06 - Windows vs Linux Benchmarks
05:16 - Feral Gamemode
05:58 - XanMod Kernel
06:11 - Xorg vs Wayland
06:20 - Esync vs Fsync
07:57 - Desktop Environments
08:22 - Conclusion

----------------------------------------------------------------------------------

Description Tags:

Optimize Linux Gaming Performance,windows vs linux,linux gaming,gaming on linux,linux vs windows gaming,how to improve linux performance,how to optimize linux for gaming,gaming windows vs linux,linux vs windows gaming performance, michael horn

----------------------------------------------------------------------------------

#linux #gaming #opensource

[aniketfuryrocks]

which nvidia driver version did you use ? nvidia 515 is better. For me steam version of games work better. Apex gets +10 fps, cs go +fps, and lower stutter than nvidia

[gespinozae]

I criticized one of your previous videos because everything you said was without data to support it. I'm really surprised you decided to do a new one showing actual data.

This is so much better, my appreciation of you and your channel has improved tremendously. Good job!

[georgepetrakis]

To anyone who wants to game on Linux: Never buy again Nvidia's GPUs. Nvidia's GPUs are not necessary better than AMD's GPUS. Nvidia have infinite cash and developers, so in collaboration with game devs they generally get better performance on Windows, because windows games are generally better optimized for Nvidia GPUs (many game companies are on Nvidia's payroll). But on Linux, the quality of the hardware shines, therefore AMD's GPUs are much better. Also AMD's open source drivers allow for better kernel integration which results in better gaming support. Let Nvidia bleed and buy ONLY AMD GPUs for linux gaming. My 2 cents.

[speedyfox]

Tip from a nerd: If you are such a nerd as I am configure your kernel. It is hard, but worth it for performance. I don't how it affects gaming though...

[ZaberfangX]

I had a Nvidia 1070 ti, I moved over to the AMD 6700XT, and I got about 3x in up lift on Linux games. I know not everyone can do that, but if people planing trying to move over Linux I would go with AMD for it's video card.

[hello_there]

From the benchmarks I've done, the tkg-kernel did significantly better than xanmod and liquorix. More than 5% higher avg fps compared to the vanilla kernel in some cases, but depends on the game. So if you are gonna use a custom kernel, try the tkg-kernel.

[overture]

Might be worth trying to see if you can measure frame times rather than FPS, to see if there really is an issue with micro stuttering and/or frame pacing?

[DMSBrian]

There is actually a *major* difference between running Xorg and Xwayland. Nvidia is known to cause heavy stutters in videos and games when using Xwayland because it's not able to synchronize application rendering (which uses Xorg and runs just fine, hence good FPS metrics), with the actual presentation, which causes visual corruption, this is admitted by Nvidia themselves. I honestly don't know why Fedora made Wayland the default for Nvidia users when it clearly wasn't ready. Since you use Nvidia, you should really run such benchmarks on Xorg because Xwayland is pretty much unsupported for those purposes at this point.

[SinistralEpoch]

One of the biggest performance boosts I've gotten as well, has been installing Gamescope. It's night and day smoothness.

[Peanutfiendsblog]

It's gotta be the GPU, my rx6600xt runs games so great on my Manjaro pc. Lots of them run better than in windows. I used to have an nvidia gpu and always had to dual boot because so many games just did weird stuff in linux for me. Weird artifacts, black bars, etc.

[AlexD-X]

Not sure it will help, but I've run with both NVidia & AMD GPUS. Currently running a 6800xt with Wayland & the same processor in the video, I get good performance on most games at 1440p (especially with AMD FSR) with nearly all those tweaks mentioned. I *think* the problem I have is refresh rate as I run a freesync monitor at 144ghz. Some games look and run great as good or better than Windows. Others (like Skyrim & Fallout 4) have high FPS but seem 'slow' which I think is refresh. But I'm a Linux noob so.... Enabling Vsnc and setting FPS max to 140 helped a lot.

Oh and I'm using Fedora and Pop OS with the kernel you mentioned, both behave differently with certain games (better, worse, not at all).

Don't forget Proton GE though....that man is a saint!

[Josephu_]

Man Its amazing to see you grow so fast much deserved man!

[julyort]

tweaks on linux:

- Use proton ge
- Gamemode
- Use variable shade rate 2x1
- Try RADV_PERFTEST=nggc in environment file (amd only)
- dxvk_async=1

[KC_rocka]

Using proton-ge over proton helps a lot with quite a few games as it adds a bunch of performance tweaks and some games do seem to run a bit better on Linux such as Elden Ring and Apex Legends, but yeah it's not quite upto Windows fps numbers and overall smoothness with some games but it's getting there.

[dexgaming]

The performance difference might be caused by the shaders not being fully compiled yet. The proton compatibility layer pre-compiles the shaders, and once it's compiled, the game no longer stutters.

[fredrik]

Fedora is pretty slow for gaming tho, changing desktop enviroment will not make mutch of a difference, however switching distro can make a big difference.

[durschfalltv]

I would like to see one with a amd GPU.

[baumstamp]

the issue with all these zen, liquorix etc kernel optimizations is, that 3d gaming perf of a "PC" is mainly affected by two things. the fact that your CPU is *TRULY* idling when doing nothing on your rig. meaning you do not waste cycles for any background tools or realtime antivirus etc.
so ideally you can watch your htop /taskmanager output for minutes and basically see all cores at 0-2% ... and maybe at MAX have a 5% spike.
also you really want to DECREASE mouse hz to 500 because anything above will introduce CPU cycles wasted.
in short, if your CPU is "as idle as possible" it will be able to push the most amount of drawcalls for the GPU, once a game is loaded up (given that you do not load any other programs in parallel)
that was part 1. now part #2 is simply that your GPU is able to reach its optimum (+ maximum) boost core so it will be fully utilized ( at least as long as your CPU can keep up).
power limits, too low or too high core voltages and old drivers can play a role here. so make sure to test through many scenarios what the optimal operating parameters for your setup are.
if you make sure that these 2 main pillars of performance are working then you can really stop wasting time in all that tweaking jungle because most of that stuff is really snake oil or at least only affects very niche setups (and as stated in video: OLDER hardware)

[anandu]

Thanks for being honest in your opinion

[Durayne]

The last benchmarks I saw,
for nvidia cards seem to have a disadvantage on wayland atm.

What apparently helps a lot is changing the kernel (especially kernel scheduler).

For gaming purposes I currently use Arch KDE-Plasma in a pretty minimalistic way (not the whole DE with all packages) with own compiled kernel (linux tkg pds) march native and -O3 / lto, with modprobed-db.
Proton-ge also compiled also with march=native and -03.

And yeah I know the -03 probably has very little impact (if you look at the phoronix benchmarks) but hey ... I can xD.
But I wanted to do a combined measurement once.

As for FPS in Witcher, I am pretty much where I am at Windows (RTX 2080 I5 9600k 16GB Ram 1TB nvme ssd),

Gamemode is also pretty adjustable btw.