powerd and increase in energy need

Kevin Oberman kob6558 at gmail.com
Fri Mar 23 06:43:20 UTC 2012

On Thu, Mar 22, 2012 at 7:26 AM,  <perryh at pluto.rain.com> wrote:
> Kevin Oberman <kob6558 at gmail.com> wrote:
>> Throttling ... is intended for thermal control, not power
>> management. The power savings will be negligible ...
> How can it possibly provide any thermal benefit, if it does not
> reduce power consumption?  Is there some significant heat source,
> other than power consumed, that throttling reduces?

It does not provide reduced power because it was designed to control
overheating. If hte CPU does not exceed the PSV temperature, it should
not have any effect at all. That is its only purpose.

If the system is idle, it makes no difference. If the CPU is loaded,
it significantly lowers power consumption, but the operation takes
longer to complete, so the total power consumed is often greater than
it would have been with no throttling. Again, TCC is for thermal
management, not power reduction.

As to report I have seen that Cx states make things worse, I simply am
baffled. I wonder if the power readings are really accurate.
Theoretically the worst possible case is that there is no advantage to
enabling Cx states. There should be no possible way to have it use
more power. This is a real possibility, too, as it is very possible to
have a system that simply would not use deeper sleep states. USB used
to do exactly that, but it's been fixed with the new USB stack in 8.
Other things like various forms of polling can also have this effect.

You can check on whether your system is ever using deeper sleep by
looking at dev.cpu.%d.cx_usage.

Finally, all studies of power consumption agree that the lowest power
usage is when CPU intensive code run as fast as possible when it is
computing and then let deeper sleep modes sharply reduce power
consumption when CPU is not needed.
R. Kevin Oberman, Network Engineer
E-mail: kob6558 at gmail.com

More information about the freebsd-stable mailing list