Kernel 6.11 ignores fixed CPU frequency

Now the issue is back in Linux 6.11.4-2-cachyos!
No issue with Linux 6.11.4-arch1-1

This issue is not solved!

Update: amd_pstate in active mode:
If I disable Boost, than the fixed CPU Frequency is locked.
If I enable boost the issue persists…some cores go beyond (boost) higher than locked Frequency.

https://archive.cachyos.org/kernel/frequency-fix-v2/

Please try this kernel

With the Linux 6.12.0-rc3-1-cachyos-rc kernel, no active Wi-Fi connections are displayed. Otherwise, the kernel runs flawlessly on my system.

no problem for me with Linux 6.12.0-rc3-1-cachyos-rc

i m suppose it’s cpupower who break something not kernel itself
edit: now bluetooth crash

With this Kernel Cores don’t go beyond, but below fixed CPU Freq.
amd_pstate in active mode

With amd_pstate in passive mode it’s the old issue, cores still go beyond.

Does this happen when booting with amd_pstate=passive directly? Or did you boot with amd_pstate=active first then switch to the passive driver during runtime?

Issue with passive mode in both cases.

@axra Has this been fixed in the latest rc kernel? (6.12-rc5)

If not, I will build kernels for you to help bisect the problem and report it to AMD.

Issue still persists in 6.12-rc5. Working only if I disable boost.

https://share.cachyos.org/znver4/kernel/amd-pstate/

We will begin bisecting now. Please try this kernel. Download the necessary packages (kernel + header) and nvidia if you use nvidia. After that, install it to your system with sudo pacman -U /path/to/pkg-1 /path/to/pkg-2

Let me know if the issue persists or not.

Same issue. Even if I disable boost.

1 Like

New kernels are up. Nginx Directory Please try these too.

Seems to work now…amd_pstate in active or passive mode…boost enabled or disabled… :+1:

Great, we’ve narrowed it down to 3 commits now. Can you please try this kernel next? Nginx Directory

Issue persists, even with boost disabled

Nginx Directory Can you try this one now? I think I’ve narrowed down the issue so I’m going to build one more kernel with the full amd-pstate patchset with the offending commit reverted.

Issue persists, even with boost disabled

Well, looks like I can’t really revert the assumed offending commit without changing a bunch of other stuff, so I’m going to continue with the bisect. This should still be the last kernel to test tho. I’ll follow up in 30 minutes.

@axra Nginx Directory What should hopefully be the last kernel is up now. Please test this. Thank you!