Battery Life < Expectations [SOLVED]

Mardoct · 2009-10-01 02:23:57

Name says it all.

Already followed any relelvant suggestions on http://www.lesswatts.org/ , but still have underwhelming battery life. < 2hrs.

It has an nvidia GPU and AMD processor if there's anything specific for power to those.

Last edited by Mardoct (2009-10-01 21:07:48)

keenerd · 2009-10-01 03:05:43

AMD means no powertop, so everything is a little more difficult to measure. And you'll have to take a lot of measurements while doing a lot of experimentation.

You can get your current (pun!) consumption with:
watch -n 1 cat /proc/acpi/battery/BAT1/state
Just make sure you are not plugged in while metering.

Physically remove things you don't use. Do you use your cd-rom? It sucks some power just sitting there. I usually pull mine out, but cdrom drives are less modular in modern laptops.

Scale back to bare minimums. Do you really need your wifi card belting out at 100% power? Play with the txpower, see what you can get away with.

Are you frequency scaling your CPU? Drop your frequency by 30% and power is cut in half. See if you can live at this frequency permanently.

Do you have a hard drive? SSDs are better about power use. I take it to an extreme and have replaced my hard drive with a microSDHC card.

How good is your heatsink? By replacing a thick heat transfer pad with a block of copper and some Arctic Silver, my average temperature dropped 10C and the fan never comes on, saving 0.5 watts.

I will never give up my massive amount of ram, but it does add up. Every 100MB of cheap desktop ram might need 2 watts of power. Quality laptop ram can be ten times more efficient, but all ram needs juice. Every bit is refreshed hundreds of times a second, even if it is not being used, and that extra gig is another watt or two.

In short, the easiest way to save energy is to do less.

edit: I forgot one. Laptops use LCD screens, which operate by obstructing white light from the backlight. It actually takes energy to display a black pixel, because the natural rest state of an LCD is transparent. Switching from an all black theme to an all white theme saved another half a watt.

Last edited by keenerd (2009-10-01 03:50:39)

elliott · 2009-10-01 04:40:09

keenerd wrote:

AMD means no powertop, so everything is a little more difficult to measure. And you'll have to take a lot of measurements while doing a lot of experimentation.

PowerTOP works just fine with AMD processors.

iphitus · 2009-10-01 06:19:47

Try laptop-mode-tools and make sure you have cpufreq setup.

keenerd · 2009-10-01 13:18:25

elliott wrote:

PowerTOP works just fine with AMD processors.

I stand corrected. And it works with Via processors as well. I had trouble with it when it was first released, and never checked up on it again. Sorry about that.

arkham · 2009-10-01 13:24:09

Btw, is it a new or an old laptop? Most of the times, after 3-4 years the battery life degrades significantly. Also, how is the performance with other operative systems?

Mardoct · 2009-10-01 18:51:40

Thanks for the tips. This laptop is only a few months old, and it's a fairly new model. I should be getting about 4 hrs. life realistically; which I was getting with the Vista it came with. This is my first time using any Linux on a watt budget, so I wasn't sure what to try.

Mardoct · 2009-10-01 21:09:31

Have CPU scaling going and laptop mode going. Still not the greatest lifetime predicted, but it's tolerable, and batterymon probably isn't reporting it very well as it has only been on for about 15 minutes.

Everyone who posted here gets an internet cookie.

crocowhile · 2009-10-24 18:46:22

I am using the settings that you find in this thread: http://ubuntuforums.org/showthread.php?t=1215928 - I don't like latop-mode-tools; they are not very efficient and I prefer setting up stuff the way I want.

The problem is that something is wrong with the new kernels. 2.2.28 in ubuntu was giving me a fairly good battery life, when tweaked properly. 2.6.31 and 2.6.32 in arch suck, due to some trouble with hr_timer. There is a bug filed here: http://bugzilla.kernel.org/show_bug.cgi?id=14424

Basically, when I run powertop I see that the kernel itself is responsible for more half of the timer calls. Firefox also seems to do it.
I hope this issue get solved.

.:B:. · 2009-10-24 18:58:51

The bug report mentions hyperthreading etc., which should only affect Intel CPUs, but the error message itself is about the high resolution timer... I haven't really checked how my 2.6.31 build handles this, I should check.

crocowhile · 2009-10-24 19:16:33

Yes, someone say their problem disappear when HT is disableb via bios. I have an Intel Duo (MBP) so, no HT should be running on my system - still I get awful numbers:

Wakeups-from-idle per second : 582.8 interval: 15.0s
Power usage (ACPI estimate): 13.8W (1.3 hours)
Top causes for wakeups:
51.1% (539.3) <kernel core> : hrtimer_start_range_ns (tick_sched_timer)
11.8% (124.9) firefox : hrtimer_start_range_ns (hrtimer_wakeup)
10.4% (110.0) <kernel IPI> : Rescheduling interrupts
5.9% ( 62.0) <interrupt> : extra timer interrupt
4.4% ( 46.9) <interrupt> : HDA Intel
2.2% ( 23.1) cpufreqd : __mod_timer (process_timeout)
1.9% ( 20.0) <interrupt> : ehci_hcd:usb2
1.9% ( 19.9) <interrupt> : eth0

crocowhile · 2009-10-26 20:31:15

I (partially) solved compiling the kernel with CONFIG_HZ=100 (it was 1000 before). Beats me why this should work considering that the kernel is tickless anyway. <kernel core> : hrtimer_start_range_ns (tick_sched_timer) goes down to 90-100. Still high but better.

al3k · 2010-01-18 03:32:27

I have been scouring the mailing lists for a few days now and haven't found anything besides recognition that there is a problem and two hints/temporary workarounds. The only thing I have seen to fix this so far is either disabling hyper threading or compiling a kernel without dyn ticks enabled. I am currently compiling a new kernel (2.6.32.3 vanilla) with dyn ticks off so I will update with how that goes. I am on a macbook pro, which doesnt support hyper threading as far as im aware, and even if it did I have no bios to disable it in anyways. Has anyone else had any luck with eliminating this issue? Its annoying only having my battery last for an hour and I hate booting into OS X

Arch Linux

#1 2009-10-01 02:23:57

Battery Life < Expectations [SOLVED]

#2 2009-10-01 03:05:43

Re: Battery Life < Expectations [SOLVED]

#3 2009-10-01 04:40:09

Re: Battery Life < Expectations [SOLVED]

#4 2009-10-01 06:19:47

Re: Battery Life < Expectations [SOLVED]

#5 2009-10-01 13:18:25

Re: Battery Life < Expectations [SOLVED]

#6 2009-10-01 13:24:09

Re: Battery Life < Expectations [SOLVED]

#7 2009-10-01 18:51:40

Re: Battery Life < Expectations [SOLVED]

#8 2009-10-01 21:09:31

Re: Battery Life < Expectations [SOLVED]

#9 2009-10-24 18:46:22

Re: Battery Life < Expectations [SOLVED]

#10 2009-10-24 18:58:51

Re: Battery Life < Expectations [SOLVED]

#11 2009-10-24 19:16:33

Re: Battery Life < Expectations [SOLVED]

#12 2009-10-26 20:31:15

Re: Battery Life < Expectations [SOLVED]

#13 2010-01-18 03:32:27

Re: Battery Life < Expectations [SOLVED]

Board footer