Kernel 3.6.2 Power Regressions (Sandy Bridge)

Svenstaro · 2013-05-27 10:01:54

Why a bug report?

firewalker · 2013-05-27 11:02:25

Obviously there is something wrong? There is a kernel version that works nicely and the next one does not.

Svenstaro · 2013-05-27 20:46:00

3.9.4 is the most recent stable kernel version and it appears to work as intended there.

stupidus · 2013-05-28 15:06:13

Power consumption is Ok with 3.9.4, just as with all the other 3.9.x. I haven't encountered the suspend/resume bug yet with 3.9.4, but only longterm observation will show if it is really gone, considering how random the bug is/was.

cris9288 · 2013-05-28 18:30:24

I've experienced suspend/resume bug three times already since upgrading to 3.9.4, the difference is that my temp seems to stop at around 55 C. Normally it would climb up to ~75 C.

firewalker · 2013-05-28 21:26:20

Yep, the issue is definitely still present with 3.9.4. +10 degrees C after resume.

firewalker · 2013-05-28 21:29:34

I also noticed that when this hapen the CPU frequency is always at max (2.3 GHz for me).

Please try

watch grep \"cpu MHz\" /proc/cpuinfo

firewalker · 2013-05-28 21:45:31

There seems to be a relative bug report.

https://bugzilla.kernel.org/show_bug.cgi?id=58801

source · 2013-05-29 16:23:53

Hi, i have Dell Inspiron N5110 with Intel® Core™ i7-2670QM CPU @ 2.20GHz × 8. I also have that problem with high temperatures after suspend. But I am interested what are your temperatures after first boot in idle. My are about 54-58 degrees C. Is that normal, are that is because of problem with kernel?

firewalker · 2013-05-29 16:47:52

This is frustrating. Even with a kernel build with CONFIG_X86_INTEL_PSTATE disabled the issue remains.

When the system start my temperature is 47~49 C. After suspend to RAM 57~61. Intel(R) Core(TM) i3-2350M CPU @ 2.30GHz.

The only viable solution for me is 3.8.11

Last edited by firewalker (2013-05-29 17:42:47)

donniezazen · 2013-05-29 17:04:26

@source I get the same temperature. I think it is more than it should but many things are at play here in kernel space and user space.

WonderWoofy · 2013-05-29 17:24:40

donniezazen wrote:

@source I get the same temperature. I think it is more than it should but many things are at play here in kernel space and user space.

@source, you may have higher temps because the system has a harder time regulating a quad core... though this *shouldn't* be the case. For the sake of a comparative reference, I am running and IVB Core i5 3210M (dual-core) with a patched kernel to enable intel_pstate on my machine. I am then also running thermald with the msr module loaded (you can just use the tarball I posted in the comments of thermald-git). Here is what I've got ATM, and this does not differ from what is typical when the system is not under load:

% sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:        +29.8°C  (crit = +100.0°C)

thinkpad-isa-0000
Adapter: ISA adapter
fan1:           0 RPM

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +37.0°C  (high = +87.0°C, crit = +105.0°C)
Core 0:         +36.0°C  (high = +87.0°C, crit = +105.0°C)
Core 1:         +36.0°C  (high = +87.0°C, crit = +105.0°C)

donniezazen · 2013-05-29 18:51:13

@WonderWoofy I really hate to see picture like those. My system never runs below 55-60 even when I have nothing resource intensive running on it. I have a thinkpad t420i 8GB RAM 2.7GHz i7-2620M Intel processor with 1GB Nvidia NVS4200 graphic card.

KaiSforza · 2013-05-29 19:00:11

@donniezazen: You have the graphics card disabled, correct? And your thinkfan settings are all up to snuff?

What I'm trying to say is that you're sure that the increased temps are because of cpu/gpu scaling issues and not just from some rogue process?

donniezazen · 2013-05-29 19:27:20

Yes, I do have everything set up like bumblebee, tlp, acpid, thinkfan, etc. I was just being envious of how people have such a low temperature system and my system runs very hot. I would say 55C is my base temperature. After suspend it goes to 80-90C.

WonderWoofy · 2013-05-29 20:22:23

@donniezazen
You know that thermald (Linux Thermal Daemon) is meant to be used with the intel_pstate driver. It keeps your system (the cpu at least) cool via p-states, t-states, and the powerclamp driver. So basically it uses the available resources found within the kernel in order to try to maintain a cool system. But in the event that there is something that spikes the temps, it also have the control of the fan in order to help it maintain.

Have you tried this? I'm not sure if it will conflict with the thinkfan daemon or not, as then you might have two things trying to control the fan at the same time.

Also, my intent wasn't to make you fell jealous of my system's temps. I just wanted to give you a point of reference at what I think a normally functioning system would yeild in terms of temps w/o system load.

source · 2013-05-29 21:01:27

Here is my temperatures

sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:        +55.5 C  (crit = +99.0 C)

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +58.0 C  (high = +86.0 C, crit = +100.0 C)
Core 0:         +58.0 C  (high = +86.0 C, crit = +100.0 C)
Core 1:         +57.0 C  (high = +86.0 C, crit = +100.0 C)
Core 2:         +55.0 C  (high = +86.0 C, crit = +100.0 C)
Core 3:         +53.0 C  (high = +86.0 C, crit = +100.0 C)

donniezazen · 2013-05-30 00:20:35

@WonderWoofy Do you have any other power management tools installed like tlp or laptop-mode-tools. I have tlp, tlp-rdw, acpid, thinkfan, etc. installed. I am giving thermald-git a try after disabling tlp and thinkfan, so, there is no conflict.

cfr · 2013-05-30 00:24:14

For another reference point, I have never seen temperatures as low as WonderWoofy's. Lowest temps tend to be high 40s. More usual is mid to high 50s when running.

$ sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:        +54.0°C  (crit = +100.0°C)

thinkpad-isa-0000
Adapter: ISA adapter
fan1:         562 RPM
temp1:        +54.0°C  
temp2:         +0.0°C  
temp3:        +54.0°C  
temp4:         +0.0°C  
temp5:         +0.0°C  
temp6:         +0.0°C  
temp7:        +27.0°C  
temp8:         +0.0°C  

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +51.0°C  (high = +86.0°C, crit = +100.0°C)
Core 0:         +52.0°C  (high = +86.0°C, crit = +100.0°C)
Core 1:         +48.0°C  (high = +86.0°C, crit = +100.0°C)

WonderWoofy · 2013-05-30 00:33:33

donniezazen wrote:

@WonderWoofy Do you have any other power management tools installed like tlp or laptop-mode-tools. I have tlp, tlp-rdw, acpid, thinkfan, etc. installed. I am giving thermald-git a try after disabling tlp and thinkfan, so, there is no conflict.

I don't use any power management tools other than thermald. I do not think you need to disable tlp, since it controls things like the power handling of pci and whatnot. This has nothing to do with the cpu power management.

I do have numerous udev rules, tmpfiles.d configurations, and module options (modprobe.d) set up so that my system is set to be less power hungry overall. What laptop-mode-tools and tlp do is change various settings if the computer is plugged or unplugged from A/C. If you use your computer and don't really notice a difference in performance between its plugged and unplugged state, then these daemons are pretty worthless. On new hardware like that which you are using, the difference is not actually enough to notice in a real world situation. It might show up in benchmarks, but benchmarks are not a good example of real usage.

Since I went through the task of setting all that crap up on my system, someone went ahead and actually compiled a list of things that can be optimized, power-wise, in the wiki. It covers things from sound codecs to pci to sata power link etc. I am not sure if eveything is in here, as I had no need to actually read all the way through this page. But it is at the very least, a good place to start.

donniezazen · 2013-05-30 00:38:31

Thanks I will check it out.

donniezazen · 2013-06-04 22:23:09

@WonderWoofy Thermald won't conflict with Thinkfan.

Hi,

Thermal daemon will not use proprietary interfaces (like thinkpad FAN) without explicitly configuring path for this in configuration. 
So it should be fine to have both daemon running unless, someone also configuring path in the thermal daemon for thinkpad fan control.

Thanks,
Srinivas

WonderWoofy · 2013-06-04 22:50:58

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.

eduedix · 2013-06-04 23:00:29

what do you see in powertop, Frequency stats tab, at actual rates? even though my pc is idle, freqs are never at lowest state(usually around 2GHz). i am wondering why.

WonderWoofy · 2013-06-04 23:22:04

I think that powertop has a hard time determining the actual clock speed of the processor at any given moment. So if this is a metric you are really interested in, t would probably be wise to use i7z instead.

Arch Linux

#476 2013-05-27 10:01:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#477 2013-05-27 11:02:25

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#478 2013-05-27 20:46:00

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#479 2013-05-28 15:06:13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#480 2013-05-28 18:30:24

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#481 2013-05-28 21:26:20

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#482 2013-05-28 21:29:34

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#483 2013-05-28 21:45:31

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#484 2013-05-29 16:23:53

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#485 2013-05-29 16:47:52

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#486 2013-05-29 17:04:26

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#487 2013-05-29 17:24:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#488 2013-05-29 18:51:13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#489 2013-05-29 19:00:11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#490 2013-05-29 19:27:20

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#491 2013-05-29 20:22:23

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#492 2013-05-29 21:01:27

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#493 2013-05-30 00:20:35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#494 2013-05-30 00:24:14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#495 2013-05-30 00:33:33

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#496 2013-05-30 00:38:31

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#497 2013-06-04 22:23:09

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#498 2013-06-04 22:50:58

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#499 2013-06-04 23:00:29

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#500 2013-06-04 23:22:04

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer