On Wed, Jan 03, 2007 at 04:54:05PM +0000, Ken Moffat wrote: > I think I'm seeing a reduced sleep time with 2.6.18.2 on an ibook > G4 (7447A cpu). At first I assumed the battery was showing it's > age, but I've just run some tests and I think there _is_ heavier > battery usage. > > My test was somewhat simplistic (charge while sleeping, open, > record battery information, sleep for approx 6 hours, open, record, > calculate). It's important to sleep for a similar time on each test, > from a full charge, because the battery use reported is not linear. > > With 2.6.18.2 (compiled with gcc-4.1.1) I saw the charge decline by > 240 units (mA?) in 21551 seconds, or an average time-per-unit of > 89.8 seconds. With 2.6.17.13 (compiled with gcc-4.0.2, running on a > different partition) I saw a decline of only 193 units in 21532 > seconds (111.5 seconds per unit). > I now think that in 2.6.18.2 something did not sleep as deeply as in earlier, or later, kernels. Current kernels seem to sleep well, so I guess the occasional "thirsty" one isn't too bad. The data I accumulated follow (one run per kernel, apart from when I mistimed the sleep, but most seem consistent):
kernel gcc units seconds secs/unit 2.6.17.3 4.0.2 193 21532 111.565 2.6.18.2 4.1.1 240 21551 89.7958 2.6.19.1 4.0.2 189 21621 114.397 2.6.19.1 4.1.1 130 14441 111.085 and after more sleep 216 21506 99.5648 2.6.19.1 4.1.1 198 22348 112.869 2.6.19.2-rc1 4.1.1 190 21663 114.016 2.6.20-rc4 4.1.1 192 21715 113.099 I guess, based on when I woke it up too soon, and then put it to sleep for the remaining time, that the wake up and restarting X uses a significant amount of power. ĸen -- das eine Mal als Tragödie, das andere Mal als Farce
