powertop shows unrealistic ACPI power usage

jlindgren · 2011-11-11 23:58:55

Hi, I just checked powertop for the wattage my laptop (Toshiba Satellite A305-6916) uses while on battery. The reading is 2 watts, with an estimated 23 hours of battery remaining. This is off by a factor of ten; I happen to know the power usage is closer to 20 watts, and the XFCE battery meter shows a realistic 2 hours of battery remaining. Anyone else seeing this problem? The last time I remember using powertop was probably a few months ago; it was working fine then.

mcmilton · 2011-11-14 16:37:03

I also have the same issue on my Asus Eeepc 1215b with freshly installed Arch64. Powertop 1.13 from official repos shows 0.9 watts and predicts the working time about 60h. Powertop 2 from aur also shows 0.9 watts but predicts 1 minute of working time. Previously I had Ubuntu 11.04 (32) with powertop 1 and 11.10 (32) with powertop 2 and both indicated appropriate 12-14 watts depending of the system load. Both systems where all was ok were 32-bits. Is this the reason? May be powertop does not support 64 bit platform?

Last edited by mcmilton (2011-11-14 17:12:42)

David Batson · 2011-11-14 17:42:21

Same thing here on a Lenovo X220 ThinkPad.

pogeymanz · 2011-11-14 19:43:55

Same thing here on a Toshiba Satellite.

It's clearly just off by a factor of ten, and then using that incorrect number to estimate the battery life.

jon · 2011-11-16 00:00:07

Confirmed here on a Lenovo Thinkpad T410s. Kernel 3.1.1. Latest BIOS.

Cdh · 2011-11-16 05:47:42

I think it's on every machine, but acpi actually reports the data correctly. So it's just powertop and only since kernel 3.1 I think.

hendry · 2011-11-18 04:43:18

powertop seems permanently broken in Archlinux. I have the 2W issue too on a X220.

hermes · 2011-11-21 00:33:19

The same here with Acer Aspire 3820G. Kernel 3.1.1.
Value in idle is 0.7/0.9W (>90 hours remaining). With 2.6 kernel it was around 8-9W, so I'd say the 10 factor error is confirmed.
acpi reports a realistic ~7 hours remaining.
I tried powertop2 and powertop-sysfs from AUR as well, same result. Powertop2 is just more accurate, reporting 712mW but predicting 1 minute of remaining time, as already stated above.
The git repository at kernel.org is empty, btw.

fsckd · 2011-12-23 18:03:30

I've updated my laptop today after a long time (3.0.x -> 3.1.5). There is a noticeable improvement in power usage.

powertop wrote:

Power usage (ACPI estimate): 0.6W (96.2 hours) (long term: 4.9W,/12.5h)

My question is where is that 0.6 coming from and what does it mean? The long term numbers seem more real (currently dropped to 3.9W,/15.4h and shortly shot up to 8W and dropping again).

Edit: acpi -b matches longterm

Edit2: laptop is Samsung NF210

Last edited by fsckd (2011-12-23 18:19:30)

.:B:. · 2011-12-23 18:28:35

Are you using the ASPM hack? I noticed powertop going bonkers after I added that to the kernel line in my boot manager.

fsckd · 2011-12-23 18:35:29

Not that I can tell. I'm using linux from core and my Grub menu.lst line is 'kernel /boot/vmlinuz-linux root//dev/disk/by-uuid/<snip> ro'.

Cdh · 2011-12-23 18:54:59

Other thread with same issue (probably a change in the kernel/acpi that powertop has not been updated to deal with):
https://bbs.archlinux.org/viewtopic.php?id=130149

Last edited by Cdh (2011-12-23 18:55:41)

fsckd · 2011-12-23 19:18:50

Thanks Cdh, I had not seen that thread when I searched.

mod action: Merged in fsckd's thread (titled "powertop reports incorrect power usage?").

Renan Birck · 2011-12-24 23:57:47

This is also happens with my Dell XPS 15 w/ kernel 3.1.5, seems that the value is divided by 10, too (or I am incredibly lucky to have a laptop that uses only 2 W .

I am using acpi_call (to disable the nVidia Optimus which I don't use in Linux) and the ASPM hack, too, so this might be the problem.

hermes · 2011-12-25 00:28:16

I use acpi_call and the ASPM hack too, but the unrealistic <1W values showed up also without the latter. Only thing I noticed, the ASPM hack lowers the power consumption of ~1W (0.1, speaking in powertop language ).

Loose_Control · 2011-12-25 00:46:23

the reason for the problem: https://bugs.archlinux.org/task/26416

Keits · 2012-01-04 11:08:00

Patch for this issue: http://ompldr.org/vYzBxeg

jlindgren · 2012-01-04 14:36:36

-			rate += watts_drawn + voltage * amperes_drawn;
+			rate += watts_drawn * voltage + voltage * amperes_drawn;

What the hell? Volts * amps = watts. Why are you multiplying watts_drawn * voltage?

Last edited by jlindgren (2012-01-04 14:36:52)

Keits · 2012-01-04 18:21:32

watts_drawn contains *ONLY* current not multiplyed by voltage.
As far as i can see it's bug in initialisation.

Keits · 2012-01-04 18:23:47

Btw -- patch is not mine.
It's working, but i didn't read it well.

Duxon · 2012-01-06 12:00:41

I feel kinda sorry for this question, but...
how to patch with the linked patch? I've never done things like that.

Keits · 2012-01-06 20:26:49

http://ompldr.org/vYzN0Mw
More 'proper' version. Though i can't explain why amperes are stored in /sys/class/power_supply/<BAT>/power_now.

--
patching is simple:

1) cd <powertop_src_dir>
2) wget http://ompldr.org/vYzN0Mw -O powertop.sys.int.113.patch
3) patch -p1 < powertop.sys.int.113.patch

graysky · 2012-01-08 11:45:14

https://bugs.archlinux.org/task/27869

Thanks Keits for the solution.

Arch Linux

#1 2011-11-11 23:58:55

powertop shows unrealistic ACPI power usage

#2 2011-11-14 16:37:03

Re: powertop shows unrealistic ACPI power usage

#3 2011-11-14 17:42:21

Re: powertop shows unrealistic ACPI power usage

#4 2011-11-14 19:43:55

Re: powertop shows unrealistic ACPI power usage

#5 2011-11-16 00:00:07

Re: powertop shows unrealistic ACPI power usage

#6 2011-11-16 05:47:42

Re: powertop shows unrealistic ACPI power usage

#7 2011-11-18 04:43:18

Re: powertop shows unrealistic ACPI power usage

#8 2011-11-21 00:33:19

Re: powertop shows unrealistic ACPI power usage

#9 2011-12-23 18:03:30

Re: powertop shows unrealistic ACPI power usage

#10 2011-12-23 18:28:35

Re: powertop shows unrealistic ACPI power usage

#11 2011-12-23 18:35:29

Re: powertop shows unrealistic ACPI power usage

#12 2011-12-23 18:54:59

Re: powertop shows unrealistic ACPI power usage

#13 2011-12-23 19:18:50

Re: powertop shows unrealistic ACPI power usage

#14 2011-12-24 23:57:47

Re: powertop shows unrealistic ACPI power usage

#15 2011-12-25 00:28:16

Re: powertop shows unrealistic ACPI power usage

#16 2011-12-25 00:46:23

Re: powertop shows unrealistic ACPI power usage

#17 2012-01-04 11:08:00

Re: powertop shows unrealistic ACPI power usage

#18 2012-01-04 14:36:36

Re: powertop shows unrealistic ACPI power usage

#19 2012-01-04 18:21:32

Re: powertop shows unrealistic ACPI power usage

#20 2012-01-04 18:23:47

Re: powertop shows unrealistic ACPI power usage

#21 2012-01-06 12:00:41

Re: powertop shows unrealistic ACPI power usage

#22 2012-01-06 20:26:49

Re: powertop shows unrealistic ACPI power usage

#23 2012-01-08 11:45:14

Re: powertop shows unrealistic ACPI power usage

Board footer