Comment # 3 on bug 1039768 from Michael Paton

(In reply to Takashi Iwai from comment #2)
> Hrm, maybe it's some devices that go to D3 at PM?
> Is the similar behavior seen when you do suspend-to-RAM?  Or is it only via
> hibernation?

I had to go back to the 4.10.13-1 kernel and try this. Going to suspend and
then resuming will allow the idle frequency to go down to a nominal 800MHz. A
subsequent hibernate and resume will take the idle frequency to a nominal
500MHz. A subsequent suspend and resume will take the idle frequency back to a
nominal 800MHz.

With kernel 4.11.0-1 a reboot and then suspend and resume takes the idle
frequency to a nominal 800MHz. This kernel does not correctly resume from
hibernate, either locking up or rebooting.