[Bug 217931] amd-pstate lacks crucial features: CPU frequency and boost control

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=217931

--- Comment #77 from Miroslav Pavleski (miroslav@pavleski.net) ---
> I'm confused here on the timing of all of these events, do you mean to say
> that
the patch did nothing?  That's what it seems like at least.

All the files in the zip are after applying the patched kernel and rebooting. 
amd_pstate_report-2025-02-17_post_patch2.txt is report created after resuming
from s2idle.

I've just double checked in the source tree used for building the kernel
package. The patch is applied. 

In addition, I did some more testing after resuming from sleep. Seems now it is
not jumping above the scaling_max_freq when not stressed. 

I stress tested all 16 cores and then it ramped up towards boost freq. (above
nominal 3.3), while having scaling_max_freq set to 1.6GHz and boost 0. However
when I stopped the cpu hogging, it ramped back down idling between 400MHz and
1400MHz. Subjectively I would say the behavior is now *way better*, the CPU is
not drawing a lot of power when not stressed, under light desktop use. The fans
stay quiet.

Seems like the amd-pstate-epp has complex algorithm that takes the
scaling_max_freq as a hint, not as hard limit. The upside is that when needed,
it adapts to those needs. The downside is that this is not ideal under battery
use, as rogue processes (e.g. browser tabs) might drain battery without the
user realizing it.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.