Hi.
So, for the title question. Take the CachyOS kernel you can download from the AUR as example. Is the performance gain really worth the time investment to make everything work perfectly. Or is it barely noticeable?
I’m kinda curious about this and I never personally installed a kernel on my own. So, I’d take it as a learning experience if anything. But, I still want to know if it’s really worth the time.
Sorry if it’s a silly question. I’m not very experienced with these things… yet.
Have you read their GitHub? https://github.com/CachyOS/linux-cachyos
Or website: https://cachyos.org/
Not yet. It’s only an example. Could do zen too or whatever else.
I’m looking for an interesting experience and maybe get something out of it in term of performance.
Is it worth it in your opinion?
Watch this video. It has enough info for you to make a decision. Generally, performance boosts are random per user, per device, per hardware, per distro.
Your system may get better, or it may not.
I use the Zen kernel, because claims are it tightens up timings a bit, which I figure may help with real-time video conference recording and streaming, and other multimedia intensive tasks.
As for any perceived differences in performance? Nothing. Not a thing.
Benchmarks are probably the only place a difference, if any, could be measured.
Thanks for the reply. Seems the general vibe around this is that the differences are not really noticeable. So, I’ll set my expectations accordingly.
I’ll take more time to think about it. I -might- still do it for the joy of experimenting with something thought.
Thanks, I’ll absolutely watch it later tonight.
Installing and running the Zen kernel is a pretty trivial thing.
sudo pacman -S linux-zen linux-zen-headers
Your existing standard kernel will remain installed, and you can simply choose which to boot with on startup.
I would advise having an alternative kernel available for boot always, as it can get you out of sticky situations. For that reason I always have the LTS kernel installed (but not necessarily used) on my systems.
Oh, it’s that easy? Wow, and I was expecting a rough ride.
I guess I was getting scared over nothing.
Depends, there are probably some benchmarks out there for different work loads. You will have to see if things improve in the workloads that you care about. For me I used the zen kernel because it comes with the necessary android kernel modules that make waydroid work.
Replying to myself. Well, it’s more for Google fodder than anything in case somebody do the same thing as me.
I did install the cachyos kernel and headers. But, after reboot with the new kernel, vulkan stopped working. (Or the entire nvidia GPU, idk I got hybrid graphics here.)
What I did to fix it is simply to reinstall the drivers with “sudo pacman -S nvidia-dkms”, rebooted. Then it still didn’t work.
After some searching I found the final step which fixed my issue which was this command “sudo dracut --regenerate-all --force” then rebooted.
Now, it’s working as intended and im free to test whatever strikes my fancy. Just leaving this reply here in case somebody search for similar issue.
I have an old PC which had noticeable performance gain when running Solus, that had linux-clear patches. I installed linux-clear on Arch but there wasn’t any noticeable change.
On my newer laptop, I used linux-zen for ~3 years. I found it to be just a slight bit more responsive (than linux lts) when system was under high CPU load. It suited my use, since I was normally running compilation jobs or VMs in the background.
But for most other day to day use, linux-zen had no benefit. I use linux-lts these days since I’m not doing much of those CPU demanding tasks.
tl;dr - performance gains/losses vary. Just try the alternate kernel and see if it gives your workflow any advantage.
I’ve had same experience.
That highly depends on the definition of “performance”. Similar situation when people are talking about “stable” software/distributions.
Kernel benchmarks like those on phoronix are mostly useless because they measure “throughput”, while most custom kernels targeting the desktop are going for “responsiveness”, which usually means less “throughput”.
For example if you’re playing a game you of course care about the average framerate, but avoiding awful 1% lows can be significantly more impactfull to your experience, even if the average is somewhat lower.
Or you do something demanding in the background - e.g. encoding a video or compiling software. How responsive is the system while doing that? Does the desktop environment become a stuttering mess or would you even notice if it wasn’t for the system monitor or the fans spinning up?
Play/experiment as you like, but have the LTS kernel installed as a fall-back.
Got that from the initial installation. No worries.
In case there is enough disk space, I would install CachyOS to benefit from their whole series of kernels, performance optimizations, customized packaging and all the extra toppings.
Yeah. I tried that but had some weird issues with it. I figured EOS with a faster kernel would be the best of both world
I used to use the lts kernel. I always had issues with the performance of my VMs. Then I installed the TKG-PDS-LTS kernel which showed a significant difference in performance. But they discontinued the prebuilt kernels. Then I switched to the cachyos-bore kernel. Which gives me a similar performance.
But this will change per user so take it with a grain of salt.
This is the way. I was very recently hit by a linux kernel bug:
Having the zen, LTS, and mainline kernels installed made it fairly effortless to narrow my buggy behavior down. I’m able to stay up-to-date with the LTS kernel until the fix lands.
For my particular case, scripting, lots of videos and music, console emulation, and some older games, zen provides just a little boost. It’s tuned to be more favorable to foreground-centric ‘end user’ loads like I put on it, rather than ‘server loads’ where there are a ton of things that all need to be working in somewhat good harmony. (I pray I never again have to run a real db server on my personal machine.)
The ONLY downside, IMO, is that when doing anything that updates the kernels or drivers needed at boot, you’ve got to rebuild an new initial ramdisk image for each kernel. That can take a few minutes for each kernel.
YMMV very much. It’s relatively easy to install, try out, and see if they match your use case.
I was using Liquorix Kernel. Also easy to install. Computer felt faster. But you need to configure stuff via TLP, otherwise it runs too hot. But After configurin it was making my laptop too hot still, so I switched to Zen. Zen and Lqx use diffent power managers.