[Tfug] "Normal" ondemand governor usage?

Claude Rubinson rubinson at u.arizona.edu
Mon Nov 24 22:16:34 MST 2008


On Mon, Nov 24, 2008 at 09:55:37PM -0600, Jeremy D Rogers wrote:
> In any case, I would double check which governor is active with "cat
> /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor" (or equiv on your
> platform) to make sure.

That was my original thought too but I am indeed using the ondemand
governor.

> That said, the lag you see could depend on what you use to watch
> it. If you are relying on a monitor of some sort, there can be a
> long delay and you might only see it bump to high speed after
> sustained demand. You could try runnig some that demands some
> processing power like compiling your kernel, or try watching it with
> something quicker like "watch -n .1 cat /proc/cpuinfo" to see if it
> is actually responding quickly to demand.

Yep, that's the issue.  I had completely forgotten that the monitor I
use only updates every few seconds, so the vast majority of bursts are
simply being missed.  Upping the refresh time shows something a bit
more like what I would have expected (though I am pleasantly surprised
at how short the bursts are and how quickly the CPU returns to it's
slower state).

Thanks.

C.




More information about the tfug mailing list