Intel P-State With Linux 5.9 Adds Passive Mode With Hardware P-States

[phoronix]

Phoronix: Intel P-State With Linux 5.9 Adds Passive Mode With Hardware P-States

Merged last week to Linux 5.9 were the main set of power management updates while hitting the kernel now are some last minute power-related changes...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.9-P-State-HWP-Passive

[aphysically]

Is active mode (with powersave?) still going to be recommended/a default or are they looking to move to passive mode + schedutil? The relationship between HWP and the prior announced changes to look to default schedutil were confusing to me when I found I couldn't even use schedutil with the usual HWP configuration of the pstate driver.

[kcrudup]

Yeah, I'm confused too, and I looked at the commit a couple of times. As it is now I've switched to using "intel_pstate=disable" and the "acpi_cpufreq" governor and I'm getting better battery life vs. before when I'd had pstate enabled (but before this commit). I'll try it and see. Sadly with this year being the huge dumpster fire it is I don't get to spend much time on my laptop outside my office so don't really get to do even seat-of-the-pants measurements much anymore.

[xerom62]

Yeah, I'm confused too, and I looked at the commit a couple of times. As it is now I've switched to using "intel_pstate=disable" and the "acpi_cpufreq" governor and I'm getting better battery life vs. before when I'd had pstate enabled (but before this commit). I'll try it and see. Sadly with this year being the huge dumpster fire it is I don't get to spend much time on my laptop outside my office so don't really get to do even seat-of-the-pants measurements much anymore.

You're probably not running at turbo frequencies as much as with the intel pstate driver.

https://www.kernel.org/doc/html/v5.8/admin-guide/pm/intel_pstate.html#intel-pstate-vs-acpi-cpufreq