Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza · 2012-12-22 21:21:13

Rolinh wrote:

Question: why .tar.gz instead of .tar.xz ?
PS: 3.7-rc8 is still on your repo but... pretty useless now

Ah. Forgot to change that. I use tar.gz locally because it's faster and easier to extract and is the default with pacman-git. (EDIT: Updated them to be tar.xz's)

And I'm going to keep them all up there so that if people happen to need an old version they can get it without too much hassle. I've also got a whole folder of bisects between 3.7-rc8 and 3.7...so yeah.

Last edited by KaiSforza (2012-12-22 21:31:54)

bolshevik · 2012-12-22 21:39:54

I also have a thinkpad X220 (2.5ghz sandy bridge i5 with intel HD 3000) and have had this issue after my last update. Computer idles 20-25 degrees hotter than normal and the power regression fix has no effect. I installed the LTS kernel and will be running that for now, as it is not affected by the bug.

ravil · 2012-12-23 13:45:23

Confirming that raw 3.8-rc1 fixed power regressions (i7-2640M).

Hspasta · 2012-12-23 22:51:03

Kai, I found a problem with the linux3.8rc1, unrelated (afaik) to the power regression. The i915 GPU hangs everytime I try to run mplayer2.

error messages:

Dec 23 17:54:10 gumgum kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Dec 23 17:54:10 gumgum kernel: [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
Dec 23 17:54:10 gumgum kernel: [drm:kick_ring] *ERROR* Kicking stuck wait on render ring

Not sure where /debug/dri... is.

Last edited by Hspasta (2012-12-23 22:55:53)

Rolinh · 2012-12-23 23:06:40

I can confirm that kernel 3.8-rc1 fixes the bug for me (Intel Core i7-2620M).
I've been using it for a few hours now and my laptop temperature is still cool (around 42°).
I'm glad there is finally a proper fix.

KaiSforza · 2012-12-23 23:21:24

Hspasta wrote:

Kai, I found a problem with the linux3.8rc1, unrelated (afaik) to the power regression. The i915 GPU hangs everytime I try to run mplayer2.
error messages:
Dec 23 17:54:10 gumgum kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Dec 23 17:54:10 gumgum kernel: [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
Dec 23 17:54:10 gumgum kernel: [drm:kick_ring] *ERROR* Kicking stuck wait on render ring
Not sure where /debug/dri... is.

I haven't run into that issue at all. Send in a kernel bug report if it keeps going.

Also, debug/dri is in /sys/kernel

Last edited by KaiSforza (2012-12-23 23:34:02)

Rolinh · 2012-12-24 09:31:43

Sorry guys, I cannot confirm the fix finally. All was fine last night but this morning temperatures went as high as 95°C without any reasons and the fan acted weirdly (like slowing speed even though the temperatures were still around 85°C ...).

degeberg · 2012-12-24 14:23:25

I've been using 3.7.1-3-ck for a couple of days with no problems.

flockyrocky · 2012-12-26 03:17:02

Just tried 3.8-rc1. Sorry it doesn't seem to work for me. My laptop temperature keeps as high as 55C. While with old kernel sometimes it can be as low as 45C or even lower.
Anyone else has the same issue?

hendry · 2012-12-26 08:05:39

I seem to be hovering around 50C using 3.8.1-1-mainline-dirty and I'm happy. A *lot* better than hitting 80C.

Does anyone know a good way to plot the temperature over time?

Must be a better way to parse temperature than using

$ sensors | grep temp1
temp1: +48.0°C (crit = +99.0°C)

The Compiler · 2012-12-26 09:07:00

With 3.8-rc1-mainline my sound is sometimes broken (just "stuck" and repeating all the time), but at least this issue is gone. I'm now at 44°-48°C.

@hendry: Maybe

cat /sys/class/thermal/thermal_zone0/temp

? Depends on your machine of course.

flockyrocky · 2012-12-27 02:01:35

If you're using GNOME, there's an extension may do the trick. It's called "CPU Temperature indicator"
Or you can do it in terminal with:

watch "sensors | grep temp1"

hendry · 2012-12-27 09:02:11

Logging my systems temperature here:
http://r2d2.webconverger.org/2012-12-27/temp.csv

Using a script like so:

#!/bin/sh
echo $(date +%s) $(cat /sys/class/thermal/thermal_zone0/temp) $(uname -r) $(cat /proc/loadavg) |
ssh sg 'cat - >> /srv/www/r2d2.webconverger.org/2012-12-27/temp.csv'

mabl · 2012-12-31 09:59:53

degeberg wrote:

I've been using 3.7.1-3-ck for a couple of days with no problems.

I'm also on 3.7.1-3-ck and thought this issue was fixes. But I just had it happen again:

# cat /sys/kernel/debug/dri/0/i915_cur_delayinfo
GT_PERF_STATUS: 0x00001ac6
RPSTAT1: 0x00049a19
Render p-state ratio: 26
Render p-state VID: 198
Render p-state limit: 255
CAGF: 1300MHz
RP CUR UP EI: 1806us
RP CUR UP: 1806us
RP PREV UP: 66000us
RP CUR DOWN EI: 215765us
RP CUR DOWN: 0us
RP PREV DOWN: 0us
Lowest (RPN) frequency: 650MHz
Nominal (RP1) frequency: 650MHz
Max non-overclocked (RP0) frequency: 1300MHz

One suspend-to-ram cycle fixed the problem again.

KaiSforza · 2012-12-31 19:54:48

For the past week I have not encountered this issue once using 3.8-rc1. I don't know if these fixes will be backported to 3.7.x, but hopefully they will.

hendry · 2013-01-04 13:49:49

I think 3.6.11 might fix the problem, since my temperatures aren't as high as 3.6.10:
http://r2d2.webconverger.org/2012-12-27/temp.csv

Not sure what change fixed it in http://www.kernel.org/pub/linux/kernel/ … Log-3.6.11

marcleb · 2013-01-05 18:33:04

Hi,
maybe this could be the solution of power consumption problem:
https://bbs.archlinux.org/viewtopic.php?pid=1030190
I think it works on my asus u36sg

donniezazen · 2013-01-06 08:04:09

The problem still exist in 3.6.11 on my system.

marcleb · 2013-01-06 20:09:49

Yes, the problem still exists in 3.6.11, but this https://bbs.archlinux.org/viewtopic.php … 5#p1030495 solved it on my system.

eduedix · 2013-01-07 10:12:14

it doesnt make sense. i thought these options had no effect anymore.

flockyrocky · 2013-01-08 02:38:00

Yes, the problem still exists in 3.6.11, but this https://bbs.archlinux.org/viewtopic.php … 5#p1030495 solved it on my system.

seems working for me until now. it haven't been happening for the last 2 days, 4 reboots.

Hspasta · 2013-01-08 04:40:00

Just discovered on linux 3.7.1-2 that the overheating only happens when I suspend with the AC attached.

jase · 2013-01-08 08:12:06

marcleb wrote:

Yes, the problem still exists in 3.6.11, but this https://bbs.archlinux.org/viewtopic.php … 5#p1030495 solved it on my system.

Strange, I had something similar settings as my kernel parameters plus the ones on that post and it didn't help. Yesterday I removed those kernel parameters, blacklisted i915* and rebooted. Now temps are under control, first time in ages.

*) I tried to use my radeon gpu but apparently I'm still using intel, even though I blacklisted it.

Rolinh · 2013-01-08 18:54:36

Hspasta wrote:

Just discovered on linux 3.7.1-2 that the overheating only happens when I suspend with the AC attached.

Everything is fine for me with 3.6.11 unless I suspend. I haven't try to suspend without AC attached though. However, it used to overheat all the time with < 3.6.11 but not anymore which is a progress.

KaiSforza · 2013-01-11 12:48:49

Okay, time for some updates on the mainline stuff. (All of the 3.8 series I've used has only had the default Arch patch applied)

So I ran 3.8-rc1 without a hitch. ran smooth, good speed, nice power usage, absolutely wonderful kernel. Ran it for a little longer than a week, no issues at all. CPU/GPU scaling was all fine.

This continued into 3.8-rc2, for a while. Then I noticed one day I came back and it was happening again. CPU was at full clock speed and GPU was in full active state. While the bug I had earlier, I only had that happen twice through a whole week. Still unacceptable, but better than every suspend/resume.

I'm onto 3.8-rc3 now. Hopefully this fixes things. I've run it for more than 24 hours, and still no issues.

I'm planning to spend the next few days building a whole bunch of kernels while bisecting and reading through kernel logs and trudging through the LKML for any sign that this has been looked into and fixed.

Rolinh wrote:

Everything is fine for me with 3.6.11 unless I suspend. I haven't try to suspend without AC attached though.

This may be your power saving settings. If you have it switch to the performance governer, then it could cause these same symptoms. (I don't know if it does the same with the gpu, but it might.)

Arch Linux

#251 2012-12-22 21:21:13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#252 2012-12-22 21:39:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#253 2012-12-23 13:45:23

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#254 2012-12-23 22:51:03

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#255 2012-12-23 23:06:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#256 2012-12-23 23:21:24

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#257 2012-12-24 09:31:43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#258 2012-12-24 14:23:25

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#259 2012-12-26 03:17:02

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#260 2012-12-26 08:05:39

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#261 2012-12-26 09:07:00

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#262 2012-12-27 02:01:35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#263 2012-12-27 09:02:11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#264 2012-12-31 09:59:53

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#265 2012-12-31 19:54:48

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#266 2013-01-04 13:49:49

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#267 2013-01-05 18:33:04

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#268 2013-01-06 08:04:09

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#269 2013-01-06 20:09:49

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#270 2013-01-07 10:12:14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#271 2013-01-08 02:38:00

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#272 2013-01-08 04:40:00

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#273 2013-01-08 08:12:06

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#274 2013-01-08 18:54:36

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#275 2013-01-11 12:48:49

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer