Re: acpicpu slowing down computer [was Re: firefox slowness -- X server problem?]

[Jukka Ruohonen <jruohonen%iki.fi@localhost>]

On Thu, Aug 19, 2010 at 02:38:43PM +0200, Thomas Klausner wrote:
> Going back to a kernel from 20100702, firefox was fast again.
> I looked at the dmesg differences and found that I had enabled acpicpu
> in the meantime. I rebuilt a kernel using sources from yesterday, but
> disabling acpicpu, and firefox is still fast; with a kernel from
> yesterday and acpicpu enabled, it is slow.

I think we need a little more information than a gut feeling.

> During boot, I always get:
> aibs0: warning over limit on 'MB Temperature'
> aibs0: warning under limit on 'CHASSIS1 FAN Speed'
> aibs0: warning under limit on 'CHASSIS2 FAN Speed'
> aibs0: warning under limit on 'CHASSIS3 FAN Speed'
> aibs0: warning under limit on 'POWER FAN Speed'

This has nothing to do with acpicpu(4).

> The CPU is
> cpu0: "Intel(R) Xeon(R) CPU           W3520  @ 2.67GHz"
> but it's often misregnized by NetBSD with a higher clockrate:
> cpu0: Intel Core i7 (Nehalem) (686-class), 3207.40 MHz, id 0x106a5
> (both lines from the same cpuctl output).

Neither does this.

The only way acpicpu(4) can currently slow down your computer in a
deterministic manner is if you have something like estd(1) running and the
available frequencies change dynamically, in which case something should
appear in dmesg (if you boot with -x). (A sysctl knob-will be likely
provided for this later, but this is still work in progress; there will be a
request for testing later.)

 -Jukka.