[ltp] Battery drain in S3 (was: Question on APM vs ACPI)

Justin Mason linux-thinkpad@linux-thinkpad.org
Wed, 08 Dec 2004 11:22:41 -0800


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Joachim Schrod writes:
> >>>>> "NK" == Nathan Kurz <nate@verse.com> writes:
> 
> NK> On Tue, Dec 07, 2004 at 07:50:59AM +0100, Bob Alexander wrote:
> >> From the help I have received from this list and reading around I have 
> >> understood that:
> >> 
> >> 1) APM is much better when suspending in low battery drainage
> 
> NK> I've read that many times too, and while I'm certain it was true
> NK> at some point for some hardware, I don't know that it is now. I
> NK> just did a little measuring on my X30 running 2.6.9 and ACPI S3
> NK> suspend-to-ram, and it seems to use abou 550 mW while sleeping. 
> NK> This is about 1.25 % of battery capacity per hour, and about what
> NK> I've seen people claim for APM sleep.
> 
> 1.24% per hour == 30% per day
> 
> Using APM and a 2.4 kernel, suspend-to-ram (equivalent to S3) eats
> roughly 10% battery power per day; both on my T21 and my T41.
> 
> Such information (30% battery drain per day in S3) kept me from
> migrating to 2.6, for now. (Suspend-to-RAM on lid-close is the most
> important function in power management that I need.) OTOH, Fionn
> Behrens reported in this thread about a drain of only 5-10% per day.
> 
> thinkwiki has no information on battery drain in S3. If you follow up
> with more experience, I compile it into information for the
> PowerManagement topic there.

so what are the fears regarding 2.6?  I'm running 2.6.7 with APM, with
virtually no problems.

Here's what works:

  - screen blanking
  - APM suspend
  - APM sleep
  - speedstep

Here's the glitches:

  - sometimes the madwifi driver reports lots of "hardware resets" to the
    syslog, even after a full power-cycle, not sure why this is.

  - sometimes -- this may be graphics-card-driver-related --
    hibernate/resume doesn't come back up fully intact.  instead, the
    display shows the image of the X desktop (as it normally does from
    resume, for 1 second or so, before the X server repaints it) -- but
    the X server then dies.  (this is using XFree86, haven't tried
    XOrg yet, as this is very infrequent.)

btw I didn't realise the thinkwiki was up!   I must go and add some
data on APM...

- --j.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Exmh CVS

iD8DBQFBt1SBMJF5cimLx9ARAtO4AJ0STKrGq9hjkVln9BnKpWv0YFPx3ACeIcBt
jqKcJO1MC88u2hQLBBORlQg=
=tcp7
-----END PGP SIGNATURE-----