Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza · 2013-05-01 19:37:35

For those of you wondering, 3.9.0 has made its way into [testing] as of yesterday morning. I'd recommend trying it out if you are still having issues.

https://mailman.archlinux.org/pipermail … 24859.html

HalosGhost · 2013-05-01 19:40:01

KaiSforza wrote:

This is an issue in the kernel. So far, I haven't had any issues with 3.9{,rc} kernels. I don't really remember how prevalent it was with the 3.8 line of kernels (that was a month and a half ago for me), but it appears to have either been solved or gotten better.

This is wonderful news. I would love to reclaim some much-needed batterylife.

All the best,

-HG

donniezazen · 2013-05-06 00:38:36

I am just hit by this bug again on 3.8.11-1-ARCH

http://imgur.com/4pH4xiz
http://imgur.com/VZn9A08

KaiSforza · 2013-05-07 15:41:04

donniezazen wrote:

I am just hit by this bug again on 3.8.11-1-ARCH
http://imgur.com/4pH4xiz
http://imgur.com/VZn9A08

Have you tried 3.9.0?

donniezazen · 2013-05-07 15:43:33

@KaiSforza Not yet. Bug doesn't seem to affect every time I suspend my system. Nvidia doesn't support 3.9 yet, so, I am hanging in with 3.8.x.

donniezazen · 2013-05-08 18:36:59

Any of you guys been noticing excessive power consumption and heat. My system with Gnome and Linux-3.8 is at least 10C hotter than Ubuntu 13.04 which is suppose to bloated.

LordChaos73 · 2013-05-08 19:12:45

donniezazen wrote:

Any of you guys been noticing excessive power consumption and heat. My system with Gnome and Linux-3.8 is at least 10C hotter than Ubuntu 13.04 which is suppose to bloated.

What's the output of the sensors command (before and after a suspended session please) ? And did you hit this bug after resuming from a suspended session?

Last edited by LordChaos73 (2013-05-08 19:14:35)

donniezazen · 2013-05-08 19:21:54

No in general I think my system temperature seems to be higher than usual. I don't know if it's a Gnome 3.8 thing or kernel related. Nothing to do with suspend cycle.

KaiSforza · 2013-05-08 19:47:45

donniezazen wrote:

No in general I think my system temperature seems to be higher than usual. I don't know if it's a Gnome 3.8 thing or kernel related. Nothing to do with suspend cycle.

Are you getting the high (edit)cpu/gpu clock speeds?

Last edited by KaiSforza (2013-05-08 19:54:51)

donniezazen · 2013-05-08 20:10:42

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
800000
[donnie@donnie-laptop ~]$ sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
[sudo] password for donnie: 
2701000

Is this helpful?

KaiSforza · 2013-05-08 20:21:11

Yeah. Unless you did something weird in between the two commands, then it's still a kernel issue. Try enabling [testing] and running 3.9. They should have nvidia blobs there that are rebuilt for the kernel bump.

donniezazen · 2013-05-08 21:01:05

Okay thanks.

regressionfree · 2013-05-09 10:25:40

hi..

I've had this regression issue since 3.4 (3.5 somewhat ok) on several distros running a laptop using i3 cpu and nvidia optimus. And I'm happy to report that now on manjaro (don't shot) and running kernel 3.9.0-1 my troubles are solved. My laptop is finally back in form and silent as a puppy.

Thanks for this thread btw. First time poster but have followed it for a long time

Thorsten Reinbold · 2013-05-09 14:18:16

KaiSforza wrote:

Yeah. Unless you did something weird in between the two commands, then it's still a kernel issue. Try enabling [testing] and running 3.9. They should have nvidia blobs there that are rebuilt for the kernel bump.

I'm running 3.9 from testing. But it seems, things have changed a bit:

┌─[nordlicht@aegis][~]
└─$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
cat: /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq: No such file or directory
┌─[nordlicht@aegis][~]
└─$ cat /sys/devices/system/cpu/cpu0/cpufreq/                 
affected_cpus                scaling_available_governors
cpuinfo_cur_freq             scaling_driver
cpuinfo_max_freq             scaling_governor
cpuinfo_min_freq             scaling_max_freq
cpuinfo_transition_latency   scaling_min_freq
related_cpus                 scaling_setspeed
┌─[nordlicht@aegis][~]
└─$ cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq 
cat: /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq: Permission denied
┌─[nordlicht@aegis][~]
└─$ sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq 
782000

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
performance powersave

Ondemand is gone?

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor 
powersave

And why "powersave" is the default?

Is this good or bad?

Last edited by Thorsten Reinbold (2013-05-09 14:23:27)

KaiSforza · 2013-05-09 14:31:58

Thorsten Reinbold wrote:

KaiSforza wrote:
Yeah. Unless you did something weird in between the two commands, then it's still a kernel issue. Try enabling [testing] and running 3.9. They should have nvidia blobs there that are rebuilt for the kernel bump.
I'm running 3.9 from testing. But it seems, things have changed a bit:
<SNIP>
┌─[nordlicht@aegis][~]
└─$ sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq 
782000

This is fine (it's just below 800MHz).

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
performance powersave
Ondemand is gone?
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor 
powersave
And why "powersave" is the default?
Is this good or bad?

I don't know what is with this new governer, but I have the same thing here, and with a simple zgrep, I can see that I did enable ondemand in the kernel. So far nothing seems wrong on my end, though.

Thorsten Reinbold · 2013-05-09 16:23:58

Ok, got it. It's because of the new intel_pstate Driver from intel, which can be disabled with "intel_pstate=disable" on the Kernel line. For now, the driver only runs on sandybridge Cpu's.

cpupower frequency-info
analyzing CPU 0:
  driver: acpi-cpufreq
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency: 10.0 us.
  hardware limits: 800 MHz - 2.30 GHz
  available frequency steps: 2.30 GHz, 2.30 GHz, 1.80 GHz, 1.60 GHz, 1.40 GHz, 1.20 GHz, 1000 MHz, 800 MHz
  available cpufreq governors: ondemand, performance
  current policy: frequency should be within 800 MHz and 2.30 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency is 800 MHz (asserted by call to hardware).
  boost state support:
    Supported: yes
    Active: yes
    25500 MHz max turbo 4 active cores
    25500 MHz max turbo 3 active cores
    25500 MHz max turbo 2 active cores
    25500 MHz max turbo 1 active cores

I don't know if the intel_pstate or acpi-cpufreq are better or what the benfit of intel_pstate is. No idea. Has anyone more or better informations about this?

Thorsten Reinbold · 2013-05-09 16:26:22

Here the output with the intel_pstate driver enabled:

cpupower frequency-info
analyzing CPU 0:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency: 0.97 ms.
  hardware limits: 800 MHz - 2.90 GHz
  available cpufreq governors: performance, powersave
  current policy: frequency should be within 800 MHz and 2.90 GHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency is 782 MHz (asserted by call to hardware).
  boost state support:
    Supported: yes
    Active: yes
    25500 MHz max turbo 4 active cores
    25500 MHz max turbo 3 active cores
    25500 MHz max turbo 2 active cores
    25500 MHz max turbo 1 active cores

Last edited by Thorsten Reinbold (2013-05-09 16:36:25)

donniezazen · 2013-05-10 21:21:25

[donnie@donnie-laptop ~]$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
800000
[donnie@donnie-laptop ~]$ sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
800000

I disabled nvidia card through bios and now my system is much more cooler and no suspend problem. Overall temp drop is very nice. Constant fan really drives me crazy.

UPDATE:- Just got hit by the bug. Disabling nvidia does bring down the system temp but doesn't have effect on suspend bug. This is really unfortunate that this bug has resurfaced for me again in 3.8.11. Time to give 3.9 a try.

Last edited by donniezazen (2013-05-10 22:36:51)

cfr · 2013-05-12 14:08:31

Does anybody have any further information on the new intel driver? I just got the 3.9.2 update and so acpi-cpufreq no longer loads ("FAILED"s from systemd). I'd just like to know if the new driver is better or worse so I know whether to disable it or not .

EDIT: See https://bugzilla.kernel.org/show_bug.cgi?id=57141. I don't understand this well enough to answer my question but perhaps somebody else will!

Last edited by cfr (2013-05-12 14:22:14)

graysky · 2013-05-12 16:02:54

Thorsten Reinbold wrote:

Here the output with the intel_pstate driver enabled...

How did you enable it? 3.9.2-1-ARCH has it hard coded:

% zgrep PSTATE /proc/config.gz 
CONFIG_X86_INTEL_PSTATE=y

Thorsten Reinbold wrote:

For now, the driver only runs on sandybridge Cpu's...

This although 3 months old, seems to agree with your statement.

EDIT: https://bbs.archlinux.org/viewtopic.php?pid=1274311 shows a simple patch to allow ivybridge chips to use the pstate driver. Running it here on my i7-3770k. Also see the g+ url linked in the thread I just posted for more discussion.

Last edited by graysky (2013-05-17 19:37:46)

Druedain · 2013-05-12 18:13:17

Anyone knows how to change from powersave to performance in new intel_pstate? After upgrade to Linux 3.9 my machine is sooo slooow… Working in Firefox is painfull, having happy moments in Minecraft changed to angry-5fps-moments in Minecraft

I've got Intel SandyBridge cpu. I've got classic frequently upgraded Arch Linux 64 with systemd on board.

cfr · 2013-05-12 18:39:13

Druedain wrote:

Anyone knows how to change from powersave to performance in new intel_pstate? After upgrade to Linux 3.9 my machine is sooo slooow… Working in Firefox is painfull, having happy moments in Minecraft changed to angry-5fps-moments in Minecraft
I've got Intel SandyBridge cpu. I've got classic frequently upgraded Arch Linux 64 with systemd on board.

Not certain but if you look at the bug report I linked to, I *think* that tells you how to do what you want. Or, at least, how to effectively do the same thing.

Druedain · 2013-05-12 18:49:45

Never mind it was easier than I though. Using this https://wiki.archlinux.org/index.php/CP … h_cpupower but putting

cpupower frequency-set -g performance

instead of using options in table above.

stupidus · 2013-05-12 21:07:10

I installed 3.9.2 now, and at a first glance it seems like the bug is finally fixed. I did not test it very long though. The power consumption is about ~11 W now. I tried one suspend/resume cycle and the power consumption did not increase.

hrkfdn · 2013-05-13 14:22:24

I just built 3.9.2 and I'm still having the problem that after wakeup from suspend the CPU core temperature goes up to 70°C (in idle 40-50°C) with no CPU usage.

Anyone else encountering this? It's so incredibly annoying.. I need to suspend/wakeup again and then it's gone.

Arch Linux

#426 2013-05-01 19:37:35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#427 2013-05-01 19:40:01

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#428 2013-05-06 00:38:36

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#429 2013-05-07 15:41:04

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#430 2013-05-07 15:43:33

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#431 2013-05-08 18:36:59

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#432 2013-05-08 19:12:45

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#433 2013-05-08 19:21:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#434 2013-05-08 19:47:45

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#435 2013-05-08 20:10:42

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#436 2013-05-08 20:21:11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#437 2013-05-08 21:01:05

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#438 2013-05-09 10:25:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#439 2013-05-09 14:18:16

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#440 2013-05-09 14:31:58

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#441 2013-05-09 16:23:58

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#442 2013-05-09 16:26:22

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#443 2013-05-10 21:21:25

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#444 2013-05-12 14:08:31

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#445 2013-05-12 16:02:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#446 2013-05-12 18:13:17

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#447 2013-05-12 18:39:13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#448 2013-05-12 18:49:45

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#449 2013-05-12 21:07:10

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#450 2013-05-13 14:22:24

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer