You are not logged in.

#1 2011-08-08 14:35:26

oded
Member
From: Israel
Registered: 2009-08-06
Posts: 26

Kernel 3 power regression concerning sandy bridge chips

Hi all,
Using 2.6.39.* acpi predicted 10-10:30 hours remaining on a 100% charged battery (9 cells battery),
on 3.0 I get a reading of 6-7 hours remaining on the same charge level.

I've also noticed I don't get the same thermal readings I used to prior to upgrading,
I used to poll /sys/class/hwmon/hwmon0/temp1_input for thinkfan controlling fan speed but it ain't there no more.
Which leads me to believe maybe the ACPI readings are plain wrong?

I'm using the latest laptop-mode tools and have cpufreq switch to ondemand on battery.
My loaded modules:

 MODULES=(fuse vboxdrv thinkpad_acpi acpi-cpufreq cpufreq_ondemand)

Thanks

Edit: This issue seems to be related to recent (detailed briefly here) power regressions involving sandy bridge chips

Last edited by oded (2011-08-08 21:25:24)

Offline

#2 2011-08-08 17:22:34

alexcriss
Member
Registered: 2010-05-24
Posts: 121

Re: Kernel 3 power regression concerning sandy bridge chips

I too noticed a power demand increase when trying 3.0 from testing (when it was there).

My powertop reported more than 15 W doing nothing, while on 2.6.39 I usually have 12-13 W. I have yet to upgrade to the latest kernel from extra and this scares the hell out of me. Two-Three Watts are just too much. I really hope that the reading is faulty and those Watts doesn't exist.

For the record, using a Thinkpad 420s

Offline

#3 2011-08-08 17:51:41

ethail
Member
From: Spain
Registered: 2011-02-10
Posts: 225

Re: Kernel 3 power regression concerning sandy bridge chips

IF you both are users of sandy bridge chips, you'll "like" to know that there is some kind of power regression linked to the sandy bridge chips. There isn't much information about that, I only found some post on the ubuntu bugtrack:

https://bugs.launchpad.net/ubuntu/+sour … bug/818830

Nothing else I have found


My GitHub Page

Best Testing Repo Warning: [testing] means it can eat you hamster, catch fire and you should keep it away from children. And I'm serious here, it's not an April 1st joke.

Offline

#4 2011-08-08 17:55:04

alexcriss
Member
Registered: 2010-05-24
Posts: 121

Re: Kernel 3 power regression concerning sandy bridge chips

That seems to be it. Anyone else know whether it was reported updtream?

Gladly my pacman cache has 2.6.39 smile I'll go with ignoring an update for the first time in my life!

EDIT: Booting 3.0 with i915.i915_enable_rc6=1 get rid of the regression. Seems that I can happily live with 3.0 smile

Last edited by alexcriss (2011-08-08 21:36:45)

Offline

#5 2011-08-08 21:16:14

Med
Member
Registered: 2011-08-08
Posts: 16

Re: Kernel 3 power regression concerning sandy bridge chips

I can confirm the problem with another sandy bridge laptop. I went from ~9W idle to about ~14W which is indeed pretty scary. Trying the fix in the bug report mentioned does not work at runtime at least.

Offline

#6 2011-08-08 21:26:37

oded
Member
From: Israel
Registered: 2009-08-06
Posts: 26

Re: Kernel 3 power regression concerning sandy bridge chips

edited title and original post to indicate the problem in a more specific manner...hopefully this will help catch the attention of other relevant users

Offline

#7 2011-08-08 21:38:27

alexcriss
Member
Registered: 2010-05-24
Posts: 121

Re: Kernel 3 power regression concerning sandy bridge chips

Med wrote:

Trying the fix in the bug report mentioned does not work at runtime at least.

It seems to work for me. Booting with i915.i915_enable_rc6=1 gets the power consumption down. Does it work for anyone else?

Offline

#8 2011-08-09 00:28:31

Deemoney14
Member
Registered: 2011-07-27
Posts: 70

Re: Kernel 3 power regression concerning sandy bridge chips

I can confirm that this makes a difference.

Previously, my battery was running around 1200-1400 mW (according to the /proc/acpi/battery/BAT0/state file), now it runs around 900-1100 mW (some of which is probably not idle time).  I wouldn't have paid this any mind were it not for this post, thanks for bringing this to my attention!


This isn't the signature you're looking for...  Move along...

Offline

#9 2011-08-09 09:02:49

oded
Member
From: Israel
Registered: 2009-08-06
Posts: 26

Re: Kernel 3 power regression concerning sandy bridge chips

alexcriss wrote:

Booting with i915.i915_enable_rc6=1 gets the power consumption down. Does it work for anyone else?

Seems to work!
<- happy camper

Last edited by oded (2011-08-09 09:03:36)

Offline

#10 2011-08-09 14:06:50

grine
Member
Registered: 2010-02-09
Posts: 38

Re: Kernel 3 power regression concerning sandy bridge chips

This seems to help quite alot, thank you smile

Still getting a bit higher power drain than before but I might have missed something during reinstall.

Offline

#11 2011-08-09 19:14:51

Haptic
Member
Registered: 2009-09-03
Posts: 149

Re: Kernel 3 power regression concerning sandy bridge chips

I can also confirm this too. I went from 11-12W to 15-16W..

Edit: the i915.i915_enable_rc6=1 definitely fixed this issue.

Last edited by Haptic (2011-08-09 19:20:03)

Offline

#12 2011-08-11 21:32:38

altercation
Member
From: Seattle
Registered: 2011-05-15
Posts: 136
Website

Re: Kernel 3 power regression concerning sandy bridge chips

lenovo x220 tablet, had the same problem from 2.6.x to 3.0 (power consumption jumped from 10W to 15W avg).

that kernel boot parameter ( i915.i915_enable_rc6=1 ) did the trick. I'm actually down to 9W avg and my fan is no longer going crazy due to bad temp data. Better than 2.6.x now.

EDIT: I'm even getting regular 8W readings, with Chrome running. Great stuff.

Last edited by altercation (2011-08-11 21:33:26)


Ethan Schoonover
Precision Colors - http://ethanschoonover.com/solarized

Offline

#13 2011-09-01 15:50:29

alphazo
Member
Registered: 2009-10-20
Posts: 163

Re: Kernel 3 power regression concerning sandy bridge chips

pcie_aspm=force i915.i915_enable_rc6=1 have fixed my power issue with 3.0.3 kernel. Installed kernel 3.0.4 yesterday and it seems that my battery no longer hold the charge. Can someone confirm?

[EDIT] Nevermind, I must have used my WinXP Virtualbox VM too much this afternoon!

Last edited by alphazo (2011-09-01 16:39:08)

Offline

Board footer

Powered by FluxBB