You are not logged in.

#126 2012-11-06 02:36:54

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

So I've been testing, out a few kernels, the mainline-rc4, lts, 3.5.6, and 3.6.5. Of these four, the only ones that reliably never have this issue are the lts and 3.5.6. After using the mainline kernel (rc3 and rc4) for about a week and a half now, I can say that they work most of the time. rc4, with the patch, reliably fails on boot, but after every subsequent suspend/resume results in a fine system. While still an issue, I have not had to worry about this for almost a whole day (I rarely restart my system, usually only suspend, which does not break the frequency scaling)

If anyone finds or writes patches that fix this on boot, please link it here so that everyone can benefit.

EDIT: I have actually noticed power savings when the scaling works. *hoping for fix...*

Last edited by KaiSforza (2012-11-06 02:46:24)


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#127 2012-11-06 10:29:53

LordChaos73
Member
From: .nl
Registered: 2008-11-11
Posts: 183

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Unfortunately, I can't fully confirm the issues others are reporting. I use Arch on the following laptop:

Sony Vaio VPCSB36FG/B
Intel® Core™ i5-2430M Processor
AMD Radeon™ HD 6470M (SPEED MODE) / Intel® HD Graphics 3000

When booting the machine and disconnecting AC, the machine will use around 9-12W when idling (with both kernels I tested). After resuming from suspend, power usage when idle will be around 14-18W, again with both kernels I tested.
I've tested both the 3.5.6 and the 3.6.5 kernel. I may be hitting another issue, because I'm also not seeing the frequency scaling issue.

EDIT: I do will do some more tests tonight to see if I can replicate the cpufreq behaviour.

Seems like this is the related kernel bug:

https://bugzilla.kernel.org/show_bug.cgi?id=48721

Last edited by LordChaos73 (2012-11-06 13:01:41)

Offline

#128 2012-11-06 14:39:38

Sander
Member
Registered: 2006-02-26
Posts: 138

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

My laptop has also been running relatively hot with 3.6.4 (I went straight from 3.3 to 3.6 because I hadn't updated in a while). It was running at 63C idle just now. I just downgraded to 3.5.6 and rebooted, and now it's at 50C. I haven't had time to test other kernel versions (3.6.6 etc). I think I'll wait it out until this issue is fixed, don't have much time for tinkering right now smile

specs:
ASUS N53SV-S1886V laptop
Intel(R) Core(TM) i7-2670QM CPU @ 2.20GHz (sandy bridge)
Intel HD Graphics 3000 / Nvidia Geforce GT 540M, nvidia driver, using bumblebee (optirun) to use it for CUDA applications
previous kernel version: 3.6.4
current kernel version: 3.5.6


You like cheese? You like peas? You'll love cheezy peas!

Offline

#129 2012-11-06 17:11:50

shadyabhi
Member
From: Bangalore
Registered: 2010-05-23
Posts: 262
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

3.6.6 - Thinkpad 220. Everything went fine after the reboot. But, then after a few hours, did a reboot again & now I can observe the issue again. Really dissapointing.

Offline

#130 2012-11-06 22:37:30

bassu
Member
Registered: 2011-12-21
Posts: 90

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

No issues here. 3.6.6 - Thinkpad X230. Idle power consumption at ~8-10W. Temps at ~44-47 °C

Last edited by bassu (2012-11-06 22:37:50)


The greatest threat to knowledge is not ignorance - it is the illusion of knowledge!

Offline

#131 2012-11-07 16:50:17

replax
Member
Registered: 2011-02-12
Posts: 90

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I also have the issue on 3.6.6 on my x220 with the i7.
about 25w power usage, supposedly caused by the fan itself though. the cpu is not working much and runs at 800mhz with the ondemand govenor.

Offline

#132 2012-11-08 00:31:27

twistedcubic
Member
Registered: 2012-03-01
Posts: 22

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

how do i see the power usage? i remembered this use to be in powertop... but when i run powertop i don't see anything about this anymore.

Offline

#133 2012-11-08 02:08:57

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 7,130

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

You need to run powertop in calibrate mode for a bit to get the power showing. (I wondered this for ages and only figured it out a couple of weeks ago.) See the man page for details.

To avoid complaints, you need to run it for at least an hour, apparently. Otherwise powertop complains it can't find files on start up. (But it functions just fine despite the errors.)


CLI Paste | How To Ask Questions

Arch Linux | x86_64 | GPT | EFI boot | refind | stub loader | systemd | LVM2 on LUKS
Lenovo x270 | Intel(R) Core(TM) i5-7200U CPU @ 2.50GHz | Intel Wireless 8265/8275 | US keyboard w/ Euro | 512G NVMe INTEL SSDPEKKF512G7L

Offline

#134 2012-11-08 05:40:42

r0b0t
Member
From: /tmp
Registered: 2009-05-24
Posts: 505

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Check this out, 
    description: Notebook
    product: HP Pavilion dv6 Notebook PC (LC952EA#ABZ)
    vendor: Hewlett-Packard
Currently running on 3.6.6-1

Powertop displays allot of bad!
http://pastie.org/5344140 (audio 100% is this normal)?
Device stats : http://pastie.org/5344143 (again allot of 100%)
Got allot of "BAD" http://pastie.org/5344146

sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:        +49.0 C  (crit = +99.0 C)

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +51.0 C  (high = +86.0 C, crit = +100.0 C)
Core 0:         +49.0 C  (high = +86.0 C, crit = +100.0 C)
Core 1:         +50.0 C  (high = +86.0 C, crit = +100.0 C)

Normally it runs at 40

Offline

#135 2012-11-08 06:43:04

justforgetme
Member
From: Germany
Registered: 2012-04-01
Posts: 51
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I can confirm these issues on a current macbook pro (non retina but that doesn't actually make a difference).
the laptop is running approx 30% hotter, I can only estimate I have no real baseline but had to reconfigure my cooling scripts to keep it in it's comfort zone. 
 
The battery discharge has become huge, moving from a 5+hr lifetime to just over 2...

AFAIK the i915 bug is the issue and is being worked on but has anybody seen an effective implementation? I see some users have tested newer kernels but do those have a bugfix for that?


$> man life
No manual entry for `life'

$> nvm...

Offline

#136 2012-11-08 16:39:46

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 7,130

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

r0b0t wrote:

Powertop displays allot of bad!
http://pastie.org/5344140 (audio 100% is this normal)?
Device stats : http://pastie.org/5344143 (again allot of 100%)
Got allot of "BAD" http://pastie.org/5344146

Why is it suddenly labelling use of ondemand as bad? Shouldn't it say that is *good*?

Do you have it set up to tweak these things? That is, do these power-saving measures usually work automatically? What happens if you switch the tunables while in powertop?


CLI Paste | How To Ask Questions

Arch Linux | x86_64 | GPT | EFI boot | refind | stub loader | systemd | LVM2 on LUKS
Lenovo x270 | Intel(R) Core(TM) i5-7200U CPU @ 2.50GHz | Intel Wireless 8265/8275 | US keyboard w/ Euro | 512G NVMe INTEL SSDPEKKF512G7L

Offline

#137 2012-11-08 20:09:20

r0b0t
Member
From: /tmp
Registered: 2009-05-24
Posts: 505

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

cfr wrote:
r0b0t wrote:

Powertop displays allot of bad!
http://pastie.org/5344140 (audio 100% is this normal)?
Device stats : http://pastie.org/5344143 (again allot of 100%)
Got allot of "BAD" http://pastie.org/5344146

Why is it suddenly labelling use of ondemand as bad? Shouldn't it say that is *good*?

Do you have it set up to tweak these things? That is, do these power-saving measures usually work automatically? What happens if you switch the tunables while in powertop?

Yes, that suprised me as well, however yes I use them:
(from rc.local - I prefer setting this from here)

/sbin/modprobe acpi-cpufreq
/sbin/modprobe cpufreq_powersave
/sbin/modprobe cpufreq_ondemand
/sbin/modprobe cpufreq_userspace
/usr/bin/cpupower -c 0 frequency-set -g ondemand
/usr/bin/cpupower -c 1 frequency-set -g powersave -u 1.2Ghz -d 0.8Ghz
/usr/bin/cpupower -c 2 frequency-set -g powersave -u 1.2Ghz -d 0.8Ghz
/usr/bin/cpupower -c 3 frequency-set -g powersave -u 1.2Ghz -d 0.8Ghz

And the governors get loaded:

analyzing CPU 0:
  driver: acpi-cpufreq
  CPUs which run at the same hardware frequency: 0 1 2 3
  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
[b]  available cpufreq governors: userspace, powersave, ondemand, performance [/b]
  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 2.30 GHz (asserted by call to hardware).
  cpufreq stats: 2.30 GHz:3.45%, 2.30 GHz:0.19%, 1.80 GHz:0.09%, 1.60 GHz:0.11%, 1.40 GHz:0.15%, 1.20 GHz:0.18%, 1000 MHz:0.25%, 800 MHz:95.56%  (82810)
  boost state support:
    Supported: yes
    Active: yes
    2700 MHz max turbo 4 active cores
    2700 MHz max turbo 3 active cores
    2700 MHz max turbo 2 active cores
    2900 MHz max turbo 1 active cores

However ondemand performance can be improved : https://wiki.archlinux.org/index.php/CP … erformance (maybe I will apply this now).

What happens?
Really the difference is like nothing, because I'm quite sure the full fault of this is of the graphic card driver / and kernel which should be reviewed and I'm quite disapointed from the work they are doing now with the kernel as is only bug-fixing and not really introducing new features or making the available ones more reliable and focusing on what made Linux used how it is now, i.e reliability, stability, easy of use.


However there are allot of other "bad's" there which I would luuuve to fix smile Any help is welcome.

Offline

#138 2012-11-08 20:18:05

WonderWoofy
Member
From: Los Gatos, CA
Registered: 2012-05-19
Posts: 8,414

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

r0b0t, that is the runniest run-on sentence I think I have ever come across. 

Of course they are not introducing new amazing features in 3.6, as new stuffs will be included in 3.7.  I think a couple 3.7 release candidates have been made available already.  Also, I fail to understand how you are "quite sure the full fault of this is of the graphic card driver / and kernel..."  Have you tried using a previous (or newer) kernel to see if the issues go away?  To me being "quite sure" and speculation are two very different things.

Offline

#139 2012-11-09 01:05:42

twistedcubic
Member
Registered: 2012-03-01
Posts: 22

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

cfr wrote:

You need to run powertop in calibrate mode for a bit to get the power showing. (I wondered this for ages and only figured it out a couple of weeks ago.) See the man page for details.

To avoid complaints, you need to run it for at least an hour, apparently. Otherwise powertop complains it can't find files on start up. (But it functions just fine despite the errors.)


Thank you so much! It appears I should get my brightness key working first... it was changing my brightness and it made it so dark that I couldn't see anything, so i had to reboot.

Offline

#140 2012-11-09 01:24:26

Hspasta
Member
Registered: 2011-12-24
Posts: 189
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

justforgetme wrote:

I can confirm these issues on a current macbook pro (non retina but that doesn't actually make a difference).
the laptop is running approx 30% hotter, I can only estimate I have no real baseline but had to reconfigure my cooling scripts to keep it in it's comfort zone. 
 
The battery discharge has become huge, moving from a 5+hr lifetime to just over 2...

AFAIK the i915 bug is the issue and is being worked on but has anybody seen an effective implementation? I see some users have tested newer kernels but do those have a bugfix for that?

No real fix yet. 3.7rc4 does not fix it either...

However, I'm seeing that it does randomly fix itself after a few suspend/resume cycles.

Offline

#141 2012-11-09 04:33:27

r0b0t
Member
From: /tmp
Registered: 2009-05-24
Posts: 505

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

WonderWoofy wrote:

r0b0t, that is the runniest run-on sentence I think I have ever come across. 

Of course they are not introducing new amazing features in 3.6, as new stuffs will be included in 3.7.  I think a couple 3.7 release candidates have been made available already.  Also, I fail to understand how you are "quite sure the full fault of this is of the graphic card driver / and kernel..."  Have you tried using a previous (or newer) kernel to see if the issues go away?  To me being "quite sure" and speculation are two very different things.

I say that for *sure* because I'm quite an "expert" with this issue and I say it's an kernel/graphic card issue because happened different time (after kernel upgrade especially - so I'm more sure that it's the KERNEL) , it begin with an upgrade of the kernel and again different time I had to switch between the versions. Having said so, I'm sorry for hurting your linux-patriotic (and btw, I love this OS and that's why I'm saying this in order to get it better, who knows maybe some kernel devel will read this smile ) fillings but not fixing an issue like this (many new linux comers I know were leaving linux just for this reason - which is quite reasonable) shows the actual predisposition of the kernel devels for making this system more stable and more reliable. [and btw I'm checking almost every release changelog and it's only bugfixing - which in some cases introduces new bugs, it's more like microxoft style bug-fixing]

Working with the sound of the FAN and having the impression that your processor will melt is not exiting.
If you have any idea however how can I fix the "bad" notes of powertop you are more then welcome.

Last edited by r0b0t (2012-11-09 04:35:57)

Offline

#142 2012-11-09 10:54:14

jijijaco
Member
Registered: 2011-04-20
Posts: 14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hi everybody.

I'm following this thread since I updated my kernel because I have this exact bug. I had to downgrade it to 3.5 because I currently have no time for testing it, I really need my laptop for my work.
I have a Asus U36SD with i5 dual core processor. I also have Optimus technology.

You are saying that the bug solves itself sometimes with 3.7rc kernel but I had the same behaviour with the 3.6.4 and 3.6.5 kernel, I don't think it's specific to 3.7 . But as said previously, I didn't really go further in the analysis of this bug ;-)

Thank you for digging it !

Last edited by jijijaco (2012-11-09 10:54:40)

Offline

#143 2012-11-09 21:10:31

eduedix
Member
Registered: 2011-02-08
Posts: 35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

does anyone know whether this bug is being worked on? what is the progress of fix? i think i am gonna sell my X220, if it goes on like that.

Offline

#144 2012-11-09 21:36:52

Mindstormscreator
Member
Registered: 2012-07-01
Posts: 186

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

eduedix wrote:

does anyone know whether this bug is being worked on? what is the progress of fix? i think i am gonna sell my X220, if it goes on like that.

While I do agree this bug is pretty horrible, I think kernel 3.5 is working just fine for the time being. You're using that, right?

So I'd say hang on a bit before selling your laptop.

Offline

#145 2012-11-09 21:52:24

wes
Member
Registered: 2011-03-05
Posts: 67

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@eduedix: As was mentioned before in the thread, I think this is the relevant bug:

https://bugzilla.kernel.org/show_bug.cgi?id=48721

It doesn't look like much progress has been made in the last week.

Selling your fantastic hardware due to faulty software seems silly.  Wait -- scratch that.  How much do you want for it?? : )

Offline

#146 2012-11-10 11:53:19

eduedix
Member
Registered: 2011-02-08
Posts: 35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

well I am kind of bored of dealing with software tongue
I just want a very high resolution ultrabook, but there is none yet.

Offline

#147 2012-11-11 00:13:36

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 7,130

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

twistedcubic wrote:
cfr wrote:

You need to run powertop in calibrate mode for a bit to get the power showing. (I wondered this for ages and only figured it out a couple of weeks ago.) See the man page for details.

To avoid complaints, you need to run it for at least an hour, apparently. Otherwise powertop complains it can't find files on start up. (But it functions just fine despite the errors.)


Thank you so much! It appears I should get my brightness key working first... it was changing my brightness and it made it so dark that I couldn't see anything, so i had to reboot.

I think it needs to do that. If you wait, it will change it back eventually. I'm guessing that it is figuring out how much power different things use by a process of elimination of some sort. At least, that happened on mine, as well. I just let it get on with it for a bit. (You don't need to sit with it darkened for the hour - it only does this at the beginning.)


CLI Paste | How To Ask Questions

Arch Linux | x86_64 | GPT | EFI boot | refind | stub loader | systemd | LVM2 on LUKS
Lenovo x270 | Intel(R) Core(TM) i5-7200U CPU @ 2.50GHz | Intel Wireless 8265/8275 | US keyboard w/ Euro | 512G NVMe INTEL SSDPEKKF512G7L

Offline

#148 2012-11-15 00:13:20

Phobophobia
Member
Registered: 2012-01-24
Posts: 11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

eduedix wrote:

well I am kind of bored of dealing with software tongue
I just want a very high resolution ultrabook, but there is none yet.

How high is "very high"? There are at least a few 13" ultrabooks with 1080p screens, and I don't know of any currently produced laptops of any size that have a higher resolution than that.

Offline

#149 2012-11-15 21:30:07

LordChaos73
Member
From: .nl
Registered: 2008-11-11
Posts: 183

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Phobophobia wrote:

How high is "very high"? There are at least a few 13" ultrabooks with 1080p screens, and I don't know of any currently produced laptops of any size that have a higher resolution than that.

Have you been sleeping under a rock lately ? wink

http://www.apple.com/macbook-pro/

Offline

#150 2012-11-16 13:39:48

r0b0t
Member
From: /tmp
Registered: 2009-05-24
Posts: 505

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +71.0 C  (high = +86.0 C, crit = +100.0 C)
Core 0:         +67.0 C  (high = +86.0 C, crit = +100.0 C)
Core 1:         +71.0 C  (high = +86.0 C, crit = +100.0 C)

sad

Offline

Board footer

Powered by FluxBB