On Mon, Jul 03, 2006 at 08:43:38AM -0700, Andre Srinivasan wrote:
>
> Just installed powersave-0.12.20 and ran through testing to see if the
> disk events occured at startup, after resume, etc. Everything seems
> to be working without needing to modify powersave/events.
Good. I seemed to remember that we fixed stuff in that area, but i was not
too sure :-)
> In addition supporting the hdparm -B switch, please consider adding
> support for the -S switch (-S 1 for aggressive powersave, -S 245 for
> performance)
Well, actually you should not need -S on remotely modern drives.
-B with an argument < 128 usually puts the drive into an aggressive
powersaving mode, that lets it spin down when it deems it worthwile.
And fact is: the drives know much better when it is worthwile to spin
down than the OS or even a fixed timeout (-S 1 lets the drive spin down
after 6 seconds). Check your drive's datasheet for details, unfortunately
they are all a little bit different wrt the actually used power saving
modes corresponding to the "-B" values.
I observed the following behaviour on modern drives (a hitachi or a toshiba,
not sure, but they behave similar) with -B 1:
- i access the drive every 5 seconds: nothing happens.
- i access the drive every 15 seconds: heads are parked immediately after
the access (low power idle mode).
- i access the drive every 60 seconds: the drive is spun down immediately
after the access.
These modi adapted to the actual use pattern, so if i had been accessing
the drive every two minutes (reading a mailinglist where i actually read
every mail), it spun down, but then i switched to LKML, where i mostly read
the mails with the "D" key, and the drive stayed on and even stopped parking
the heads.
And the best argument is, that spinning down the drive does not actually save
you an significant amount of power, anyway:
- a modern drive draws about 0.55W in "low power idle" (heads parked, disk
spinning) mode.
- in standby mode, it still draws about 0.3W.
So you can save ~0.25W, but you have to account the increased startup current
for the spindle motor, too, so i dare to say that from a power saving POV,
spinning down the drive is practically worthless.
This is not only theory (i use this stuff in my everyday work, and i commute
pretty long times every week, where i need to squeeze the last bit out of the
battery :-) but i also saw this in the real world: On a Dell D600, all the
most agressive tweaking of the disk did not give me a noticable gain in
battery life (it was less than 5 minutes, nothing you could "measure" with
my totally unscientific "use it until the battery runs out" testcase). Once
the ACPI processor driver started supporting C4 states, i got ~15-20 minutes
extra suddenly, _that_ was something to be noticed easily :-)
But anyway, we won't restrict power users on their options, this is why there
are three variables
HDPARM_RAW_PERF=""
HDPARM_RAW_SAVE=""
HDPARM_RAW_AGGR=""
in /etc/powersave/disk.
Just do
HDPARM_RAW_PERF="-S 0"
HDPARM_RAW_SAVE="-S 10"
HDPARM_RAW_AGGR="-S 1"
and it should work for you.
> The bug in thinkpad_acpi_events still exists: It sets the PATH and
> then invokes wttyhx. /sbin is needed in the PATH otherwise pidof is
> not found.
I fixed this in the current svn trunk by adding ":/sbin:/usr/sbin" to
the PATH.
> One whine is that the FC5 RPM didn't create .rpmsave files.
That's a packaging issue and i have no real clue of packaging at all, so
somebody else will have to tackle this.
--
Stefan Seyfried | "Please, just tell people
QA / R&D Team Mobile Devices | to use KDE."
SUSE LINUX Products GmbH, Nürnberg | -- Linus Torvalds
_______________________________________________
powersave-devel mailing list
powersave-devel@forge.novell.com
http://forge.novell.com/mailman/listinfo/powersave-devel
