AMD idle speed higher with kernel 6.13 than with 6.12

I messed around with this over the weekend. Learned some too. I eventually found a combination of things that seems to fit me well.

I have to accept the decisions leading to the kernel patch. Those folks are much smarter than I am and I can’t change their decisions anyway. The higher clock at idle remains.

So I focussed mainly on tuned. I had been using the balanced profile. After experimentation, for my use case, the powersave profile makes most sense. Using it, my CPU idle temps (a reasonable proxy for energy use) dropped 7C to 34C, which is about 9C higher than ambient. This, despite the 2.2x higher clock.

Response times in the apps I use routinely seem no different, even though the powersave profile limits the clock to the advertised base of 3.3 GHz (mis-reported above as 3.9, sorry).

I did notice the difference when building AUR packages, but it’s easy enough to switch to the throughput-performance profile before–or even during–a build. Doing that, thermals then limit the clock. Had no issues with thermal limits on most things I tested.

However ungoogled-chromium hit the thermal limits pretty hard at max clock, then throttled back the clock while maintaining 100% usage on all 16 cores and 32 threads. Same on the laptop with a Ryzen 7 7840U (8C/16T). And that’s the first time I noticed any significant difference between the two computers.

But, since my actual routine use case barely nudges the throttle on my my third computer–an ancient Intel laptop (2C/2T @ 1.5 GHz)–let alone the beast of a desktop, or even the Framework laptop, running in powersave is perfectly fine for me in both instances. Now settled on 6.13 on the desktop, 6.14 on the lappy.

Thanks for pointing me in the right direction.

PS: There’s a similar thread over on Arch Forums, but I’m not smart enough to pass the test to create an account, so I couldn’t refer that poster to this. Perhaps you or someone else can?

1 Like