| 1 |
Alexander Meyer (ali@×××××××××××××××.de) scribbled: |
| 2 |
> On Mi, Jun 29 2005 at 01:32:29 +0200, Colin Leroy wrote: |
| 3 |
> > On Wed, 29 Jun 2005 12:27:39 +0200 |
| 4 |
> > Alexander Meyer <ali@×××××××××××××××.de> wrote: |
| 5 |
> > |
| 6 |
> > > Well, if there's no means to control when the fan starts it probably |
| 7 |
> > > _is_ something like this. i'll try some earlier 2.6 kernels then since |
| 8 |
> > > Stefan earlier stated that this change didn't occur at the 2.4->2.6 |
| 9 |
> > > switch but somewhere in the early 2.6 series (IIRC before 2.6.7). |
| 10 |
> > > i'm hardly using the machine anymore because that goddamn fan is just |
| 11 |
> > > too annoying and if i try to compensate by turning up my music the |
| 12 |
> > > neighbors will complain after a while ;( |
| 13 |
> > |
| 14 |
> > May be due to the defaut HZ value change. |
| 15 |
> |
| 16 |
> Now _that_ is strange: I recently changed that back from 1000 to 100 but |
| 17 |
> with no visible (or audible that is) effect whatsoever. |
| 18 |
> |
| 19 |
|
| 20 |
This is correct behavior. The value of HZ defines the minimum |
| 21 |
resolution of system calls like usleep() and select(). At 100Hz, the |
| 22 |
minimum resolution is ~20ms (= 2 * 1 / f) for usleep and ~10ms (= 1 / f) |
| 23 |
for select(). Setting it to 1000Hz adjusts those values down to ~2ms |
| 24 |
and ~1ms respectively. I've only found this useful with old/half-assed |
| 25 |
code that polls interfaces. They'll use usleep() to throttle the |
| 26 |
polling and not max out the CPU. |
| 27 |
|
| 28 |
But that is background info, the main point is that it has no effect on |
| 29 |
power consumption. |
| 30 |
|
| 31 |
Do your system monitors (gkrellm, top, whatever) show a maxed out cpu? |
| 32 |
|
| 33 |
Cooper. |
| 34 |
-- |
| 35 |
gentoo-ppc-user@g.o mailing list |
Archives