Kernel 3.6.2 Power Regressions (Sandy Bridge)

eduedix · 2013-06-05 15:03:37

wonderwoofy, can you share here your results of i7z please?

hrkfdn · 2013-06-05 15:46:52

Just had a "hot wakeup" with 3.9.4 again so the problem persists. This is terrible..

WonderWoofy · 2013-06-05 17:06:40

Uuhhh... it is a contantly changing realtime applcation. What is it you want me to share? My frequency normally sits around 1.7GHz, and i have an i5 3210M which has a 2.5GHz clock speed with a 3.2GHz turbo boost. The minimum speed is 1.2GHz.

What I am trying to get at is that the clock speed reported is of little importance, but powertop is not going to be very accurate anyway. What you really want to be looking at is not what the clock speed is, but rather how much your prcessor goes into power saving states. My processor for example seems to be in C7 for 98-99% of the time right now (system is mostly idle, as I'm just checking email and perusing the forums).

You have to take into consideration that although I am participating in these forums, I am not likely to suffer from the same problems as the others in this thread because my processor is not a Sandy Bridge. I have an Ivy Bridge, but I have also "patched" the kernel so that the intel_pstate driver works for my system. (I say "patched" because in reality it is just the addition of a single line of code, so I just added it by hand.) I also run Intel's Linux Thermal Daemon, which seems to handle thermal/power pretty damn well.

donniezazen · 2013-06-05 18:34:47

WonderWoofy wrote:

Though it may not conflict, I think it might be a better idea to actually configure thermald to control the fan instead. It just seems like a single central system for controlling thermal levels is likely to be vastly superior to having multiple things that really don't know about each other trying to all act in concert.

That makes sense. My reading tells me there is no obvious config file for thermal-daemon. It tries to do everything automatically or user created XML file.

firewalker · 2013-06-05 19:18:40

I have opened a bug report. Is someone has similar issues...

https://bugzilla.kernel.org/show_bug.cgi?id=58971

donniezazen · 2013-06-05 19:28:20

@firewalker In my case temperature goes to over 80C and not just an increase of 10C. I would say temperature increases to about 30C. Do you think these are related or different problems?

firewalker · 2013-06-05 19:47:58

Yes, I believe it is related. Your increase of the temperature may be bigger because your cpu has much more power than my i3.

Last edited by firewalker (2013-06-05 19:49:22)

firewalker · 2013-06-11 08:21:41

My problem is i915 related. I found the offending patch.

[15239099d7a7a9ecdc1ccb5b187ae4cda5488ff9] drm/i915: enable irqs earlier when resuming.

http://o.cs.uvic.ca:20810/perl/cid.pl?c … cda5488ff9

Removing this patch, everything goes back to normal.

hrkfdn · 2013-06-11 13:00:42

firewalker wrote:

My problem is i915 related. I found the offending patch.
[15239099d7a7a9ecdc1ccb5b187ae4cda5488ff9] drm/i915: enable irqs earlier when resuming.
http://o.cs.uvic.ca:20810/perl/cid.pl?c … cda5488ff9
Removing this patch, everything goes back to normal.

Great, thanks for letting us know. I have just reverted that patch and recompiled my kernel.

firewalker · 2013-06-11 13:09:20

Let us know if it will do the trick for you.

Do you use i915 also?

Last edited by firewalker (2013-06-11 13:14:13)

WonderWoofy · 2013-06-11 14:52:10

@firewalker, have you reported this?

firewalker · 2013-06-11 15:07:19

Yes, I have.

https://bugzilla.kernel.org/show_bug.cgi?id=58971

KaiSforza · 2013-06-11 19:12:50

The only problem I can see with this is that the commit happened after tag v3.8, and this bug affected 3.7 as well.

hrkfdn · 2013-06-11 23:33:58

firewalker wrote:

Let us know if it will do the trick for you.
Do you use i915 also?

So far so good. But the problem occured rather randomly before, so I'd probably have to wait at least a few days.

Yea, Thinkpad X220i here with the Intel HD 3000 GPU.

firewalker · 2013-06-12 00:33:14

Try

sudo cpupower frequency-set -g performance

and then suspend/resume. It always triggered the problem for me.

ayr0 · 2013-06-13 22:46:26

I seem to be experiencing a heat problem on my ThinkPad T520 with i7 sandybridge.

I can attribute the heat problem to my intel IGP running at 100% sometimes at wakeup. Happens about every third or fouth wake-from-sleep. After a reboot, all is back to normal.

running linux 3.9.5 with the latest intel driver.

dr_undecided · 2013-06-15 16:35:16

Is this fixed? Is there any tweak I can do to make the bug dissapear?
I really want to go back to linux, but this bug makes me look at other distros

donniezazen · 2013-06-15 16:44:53

It's a kernel problem distro hopping won't help unless you to back distro using to pre-3.7 kernel.

firewalker · 2013-06-15 18:10:31

hrkfdn wrote:

So far so good. But the problem occured rather randomly before, so I'd probably have to wait at least a few days.
Yea, Thinkpad X220i here with the Intel HD 3000 GPU.

Any news?

hrkfdn · 2013-06-15 19:13:36

firewalker wrote:

hrkfdn wrote:
So far so good. But the problem occured rather randomly before, so I'd probably have to wait at least a few days.
Yea, Thinkpad X220i here with the Intel HD 3000 GPU.
Any news?

I haven't tried with your command, but I still haven't encountered the bug even though I performed multiple suspend/wakeup cycles since then.

Seems to have done the trick!

I've been polling the bugreport you filed.. Still no statement from the i915 guys?

Last edited by hrkfdn (2013-06-15 19:14:25)

firewalker · 2013-06-15 20:06:19

Post to the bug report (when you make sure it's ok) that you had the same issue and that the removal of the patch fixed it for you too.

hrkfdn · 2013-06-16 00:38:14

firewalker wrote:

Post to the bug report (when you make sure it's ok) that you had the same issue and that the removal of the patch fixed it for you too.

That was too early, It just happened again Another suspend/resume cycle fixed it.

Last edited by hrkfdn (2013-06-16 00:38:38)

firewalker · 2013-06-16 10:04:51

/could it be just a fluke? Could you fix it with another suspend/resume cycle before?

hrkfdn · 2013-06-16 12:28:54

firewalker wrote:

/could it be just a fluke? Could you fix it with another suspend/resume cycle before?

Yes, that's how I used to fix it before.

donniezazen · 2013-06-17 15:57:35

firewalker wrote:

My problem is i915 related. I found the offending patch.
[15239099d7a7a9ecdc1ccb5b187ae4cda5488ff9] drm/i915: enable irqs earlier when resuming.
http://o.cs.uvic.ca:20810/perl/cid.pl?c … cda5488ff9
Removing this patch, everything goes back to normal.

Did you have to recompile the kernerl to remove the patch or is there a dynamic way to revert a patch?

My problem has aggravated since 3.9.6-1. I am booting into a system which is running at full frequency 3.4GHz and a temperature of 95-97C.

Arch Linux

#501 2013-06-05 15:03:37

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#502 2013-06-05 15:46:52

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#503 2013-06-05 17:06:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#504 2013-06-05 18:34:47

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#505 2013-06-05 19:18:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#506 2013-06-05 19:28:20

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#507 2013-06-05 19:47:58

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#508 2013-06-11 08:21:41

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#509 2013-06-11 13:00:42

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#510 2013-06-11 13:09:20

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#511 2013-06-11 14:52:10

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#512 2013-06-11 15:07:19

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#513 2013-06-11 19:12:50

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#514 2013-06-11 23:33:58

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#515 2013-06-12 00:33:14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#516 2013-06-13 22:46:26

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#517 2013-06-15 16:35:16

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#518 2013-06-15 16:44:53

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#519 2013-06-15 18:10:31

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#520 2013-06-15 19:13:36

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#521 2013-06-15 20:06:19

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#522 2013-06-16 00:38:14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#523 2013-06-16 10:04:51

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#524 2013-06-16 12:28:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#525 2013-06-17 15:57:35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer