Are Custom Linux Kernels Faster than Stock?

Are Custom Linux Kernels Faster than Stock?
Benchmarks are done and will be compared using phoronix test suite. We will be analyzing 3 kernels, Liquorix, Mainline, and Xanmod.
Timestamps:
00:00 Introduction
01:34 Xanmod Overview
04:17 Benchmarking Setup
07:15 Benchmarking Run
08:47 Benchmark Results
10:24 Conclusion
---

[peterjansen]

Chris, for the next time that you benchmark a kernel for a game, I recommend you to use lower graphics settings. It will show bigger differences but more importantly, it is easily possible that with the highest settings you get 0 difference while with lower settings and possibly a lower resolution you get relevant difference. You want the bottleneck to be at the CPU, not at the graphics card. Trust me on this, I understand hardware well. (For people who interpret this wrong, I do not claim that a CPU-review should necessarily be done with lower settings, that is a different matter for which more factors are relevant.)

[KodyVonBargen]

I'm really digging the whole timeline splitting/naming. Lets me skip to what I really want to know.

[ClifffSVK]

Windows users: "I'm waiting for the update process to finish so I can use my computer again..."
Linux users: "Hey, let's install some of these cool custom kernels to see if I can get more fps in games!"

[worldhello]

@10:19 Exactly, frame times are more important than the highest fps.

[tidzej]

Hi Chris. I think that it's essential in such tests to show lowest fps, not only highest and average.

[lperkins]

Your GPU was pegged at 100% the whole time, so the only thing you're seeing here is any slight differences in the AMDGPU kernel driver... Try a CPU bound test to see real-world kernel differences...

[SamichHunter]

I use a first gen Ryzen 8 core and have been using the Liquorix kernel for stability. I find that I get fewer random reboots and system freezes with it compaired to the mainline kernel. I have not used, or heard of, Xanmod but I will be trying it out now.
Thx Chris, another good and informative video!

[tonysu]

Your chosen benchmarking is GPU dependent, less effected by CPU.
If you use the same GPU drivers which are today generally distributed as kernel modules, of course you'll see nearly the same performance.

If anything,
This video is an example that you really have to know what you're doing to construct meaningful testing.
And, if you're going to test custom kernels or add or remove mods manually (Today's Kernel Loadable Modules can be loaded and unloaded by Users, it's only what is in the initrd you can't modify), you gotta know what's relevant or not to what you are doing.

Sorry Chris,
This was a good try but this video was likely educational in a very unintentional way.

[JessicaFEREM]

I ended up installing Xanmod for my ryzen laptop as a solution to a "CPU leak" I was having for KDE plasma.
turns out the kernel wasn't the problem, KDE had some messed up config or something that I found a command list that ended up fixing it properly.
but after installing xanmod I'll probably keep it just because it's a better kernel

[MGThePro]

You forgot tkg-pds, which is only easily installed on arch, but is IMO the best custom kernel

[PixelTrik]

I've recently switched to gentoo and the performance is great on my custom kernel already. Plus, the memory footprint of my kernel is as low as 85MB before I fired X server with DWM! I can play some windows games like blur on highest settings already but I need to explore more to see how gentoo performs in gaming.

[jimdeadlock]

I've been using the Zen kernel on Arch since you talked about it the other day. I play a lot of Rust and I can't say I've noticed any difference, but it works fine so I'll keep it anyway - they say it's better for general desktop usage as well as gaming.

[seylaw]

I am very impressed by a custom compiled Xanmod + BMQ scheduler (with some tweaked settings and more aggressive compilation flags).

[nobu]

i do use xanmod kernel and it makes a day and night difference for me. i love the project.

[AnzanHoshinRoshi]

Thank you, Chris. I understand why people would enjoy hacking and futzing but my systems have always done what I want and more.

[tamodolo]

The average FPS don't tell much as it's only GPU benchmark. No kernel was tested in your benchmark unfortunately...

For that to actualy be a kernel benchmark you must test response time speed (aka input lag) as these kernels aims to optimize latency. For that you must have some way to measure actual time of input and response on the screen. The faster, the better.

This is mainly because a lower fps but with low latency will generaly fells better to play than a high fps with higher latency.

Also, if you realy want to test how the kernel affects the performance of the CPU in games you'll also need to remove GPU bottlenecks. For a RX580 you need to run benckmarks at lower resolutions with every graphic detail set to the highest (without any AA) to enable every simulation the game uses. A 480p or even a 720p test should be enough to show that.

[hewfrebie]

Nice bootscreen, Shodan the Big Sister is watching us! XD

[onioncontrol]

I notice very little difference between the experimental kernel and using zen, but I get way better performance and compatibility on laptop using the experimental kernel over the stable kernel in my Manjaro install.

[arashkmahshidfar]

I just use the regular linux kernel since it has less compatibility problems.

Edit: I know it's not necessarily better but I use optimus manager since I have a laptop with hydro graphics (multi graphics). And it doesn't run even on zen kernel. I'm just not sure anything else would work.

[stargirl]

I personally use the TKG Kernel with the MuQSS scheduler. Even though it is only officially supported by Arch I have it compiled on openSUSE.
I was actually going to test xanmod patches on TKG (with real time patches) and see what I can come up with.
Gotta make the custom kernels even more custom c;