Intel P-State Driver Preparing To Default To Passive Mode For More Systems

Currently being tested ahead of the Linux 5.8 kernel cycle is a change so the Intel P-State CPU frequency scaling driver will begin defaulting to its passive mode for systems without hardware-managed P-States.

P-State's passive mode will become the default for more systems on Linux 5.8 if this change is not reverted. The passive mode causes P-State to behave like a conventional CPUFreq scaling driver and feeds all of the optimized configuration bits into CPUFreq, such as all of the available P-States. More details on the active vs. passive mode difference for P-State via this updated documentation.

The change testing P-State in passive mode by default for non-HWP systems is currently being vetted by the power management subsystem's "bleeding edge" kernel branch.

This change is part of the Intel P-State transition to using the Schedutil governor by default for relying upon the kernel's scheduler utilization data for making better CPU frequency scaling decisions. But as shown in initial testing, Schedutil still is generally performing worse than the performance governor at least on CPUs tested so far.