The Performance Impact From Different Arch Linux Kernel Flavors

[ferry]

Personally, I am amazed at just how well the -RT kernel held up to those tests compared to the rest. I do see where it did not do well against the Stress-NG test bench, but with the other tests....again....amazed.

Without a RT workload there should not be large difference. With a RT workload the -RT kernel will much prefer low latency over throughput. For processes with elevated priority... So slower disk, less USB bandwidth, to reduce interrupt latency from 1 - 100ms, to 20 - 50us. As a desktop top user that will not "feel" as better responsiveness. In a control loop otoh it may make the difference between keeping the car within the lane or crashing into a tree.

[Arthus]

The gaming benchmarks once again might have profited from some 1% low FPS graphs.

I think this is where a RT kernel could help: Getting input latency, frame consistency and presentation times right.

[rstrube]

Michael Thanks for the benchmarks, very useful! I'm been running Arch and my main system for the past couple of years and was always curious if there was any benefit to switching to the zen kernel. Good to see the overall it's basically a wash, unless one has some specific workflows that benefit.

[caligula]

Would a vanilla kernel with custom configuration perform better in any circumstance?

[V1tol]

Can't really recommend -zen kernel since it has some debug options enabled (that hurt performance a little). I am used to build my custom kernel using this scripts and on Arch it is as simple as firing a single console command. There are prebuilt kernels too on release page if you want to try.

[deusexmachina]

So you can basically used the hardened kernel guilt free on servers; thought so!

[avis]

Nope, Linux with proper CPU governor and Gnome is far more responsive.

Anything to back it up? No? Just what I thought. Starting with Windows Vista, the vast majority of UI rendering in Windows is done on GPU and Linux is nowhere near close.

[jrch2k8]

Anything to back it up? No? Just what I thought. Starting with Windows Vista, the vast majority of UI rendering in Windows is done on GPU and Linux is nowhere near close.

Windows the moment you put pressure on the hardware the desktop just freezes, there is no responsiveness argument.

just try to work on blender while you run aida64 for example, you will see.
e
On Arch with Gnome 43 wayland i literally trash the CPU converting videos to AV1 and compiling at the same time and unless the OOM (Dont happens often since i have 32gb+ on all my systems) hit hard but browsers and stuff still work good enough to really dont care too much about responsiveness

[avis]

Windows the moment you put pressure on the hardware the desktop just freezes, there is no responsiveness argument.

just try to work on blender while you run aida64 for example, you will see.
e
On Arch with Gnome 43 wayland i literally trash the CPU converting videos to AV1 and compiling at the same time and unless the OOM (Dont happens often since i have 32gb+ on all my systems) hit hard but browsers and stuff still work good enough to really dont care too much about responsiveness

Anything to back it up? Like recording a video or providing a reproducible test case? No? When was the last time you ran Windows?

I've run Windows since Windows 3.0. The last version of Windows which "put pressure on the hardware the desktop just freezes" was Windows 98SE back in 1999 because it wasn't a true multitasking OS. Windows NT 4.0 and 2000 were both rock solid and ran just fine under any conditions (aside from RAM pressure but Linux crumbles under RAM pressure as well). NT 4.0 wasn't popular because it didn't support a lot of HW, proper support came only with Windows 2000.

[edxposed]

How does the x86-64-v4 kernel of CachyOS perform on the 7950X?