You are not logged in.

#101 2012-11-01 00:27:43

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 5,675

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

3.6.4 seems slightly worse, if possible, on my hardware. (But I'm not sure if this is just because bluetooth powersaving isn't getting triggered by default for some reason:

Bad           Autosuspend for USB device Broadcom Bluetooth Device [Broadcom Corp]

now appears in powertop whereas before I only had two "Bad" toggles I couldn't figure out... sad


How To Ask Questions The Smart Way | Help Vampires

Arch Linux | x86_64 | GPT | EFI boot | grub2 | systemd | LVM2 on LUKS
Lenovo x121e | Intel(R) Core(TM) i3-2367M CPU @ 1.40GHz GenuineIntel | Intel Centrino Wireless-N 1000 | US keyboard with Euro | 320G 7200 RPM Seagate HDD

Offline

#102 2012-11-01 00:30:52

WonderWoofy
Member
From: Los Gatos, CA
Registered: 2012-05-19
Posts: 8,412

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@cfr, did you try using the mainline kernel as suggested a bit earlier in this thread?  I know your machine is not an x220, but it is indeed sandy bridge.  I dunno, worth a try?
As far as the "Bad" toggles, besides the bluetooth, what else do you have listed, out of curiosity?

Offline

#103 2012-11-01 01:14:15

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 5,675

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I haven't tried it. I'm a bit nervous about trying to compile a kernel and even more about using one if I do manage to compile it. I'm sure this is silly but... Also, I haven't really had the time to try to figure it out.

The other "bad" toggles I get are to do with VM writeback timeout and Audio codec power management. I know that these toggles do something like the following:
- /sys/module/snd_hda_intel/parameters/power_save should be 1 for Audio codec power management
- /proc/sys/vm/dirty_writeback_centisecs set to 1500 for VM writeback timeout
However, I'm not currently sure just what these mean, whether they should just be used on battery or what undesirable consequences they might have. I'm also unclear whether laptop mode tools can handle these or whether I need to do something else.

I'm a bit confused by the complaint about bluetooth because bluetooth is disabled when I'm on battery. (laptop mode tools handles this and powertop recognises the management of the "interface" as "good".)


How To Ask Questions The Smart Way | Help Vampires

Arch Linux | x86_64 | GPT | EFI boot | grub2 | systemd | LVM2 on LUKS
Lenovo x121e | Intel(R) Core(TM) i3-2367M CPU @ 1.40GHz GenuineIntel | Intel Centrino Wireless-N 1000 | US keyboard with Euro | 320G 7200 RPM Seagate HDD

Offline

#104 2012-11-01 01:23:17

WonderWoofy
Member
From: Los Gatos, CA
Registered: 2012-05-19
Posts: 8,412

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

As far as the snd_hda_intel, it is simply turning your sound "card" off when not in use.  The wort thing that can happen is that sometimes there is a delay for a split second when you start playing something with audio again.

Here s what I have for that... not sure if the file name is numbered right, but I figure 66 is a safe ways down the line, and it works.  It is a udev rule, so it is located in /etc/udev/rules.d/ :

% cat 66-sndpower.rules
#turn on intel sound card power saving feature (powertop recommended)
ACTION=="add", KERNEL=="snd_hda_intel", SUBSYSTEM=="module", TEST=="parameters/power_save", ATTR{parameters/power_save}="1"

As far as the dirty_writeback_centisecs, I think it has something to do with the harddrive spin down and how often it spins back up to write.  I could be totally off on this, but I know it is what laptop mode is supposed to default to. 

For this one, I use /etc/sysctl.d with a file as follows:

%  cat vm-writeback.conf
## set /proc/sys/vm/dirty_writeback_centisecs to Powertop's recommended 1500
vm.dirty_writeback_centisecs=1500

Offline

#105 2012-11-01 05:37:59

Hspasta
Member
Registered: 2011-12-24
Posts: 187
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

This issue will not go away for me...

3.6.4 still occurs, usually.
3.7rc2 without posted patch, still occurs.
3.7rc3 with patch, occurs sometimes...

I'm out of options.

On powertop, I notice that the GPU(intel) is active 100% of the time when it heats up. When its normal, its usually in RC6 ~90% of the time.

Last edited by Hspasta (2012-11-01 05:40:53)

Offline

#106 2012-11-01 05:53:02

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

3.6.4: Issue continues. Rare chance of booting into a nice power consumption

3.7-rc2/3, vanilla: Issue seems to be fixed, but a suspend will usually bring it back to 20W idle.
3.7-rc2/3, patched: Issue is 'more fixed'. It usually takes 2-3 suspends for this behavior to surface.

So far, I am still running 3.5.6. Until this issue truly gets fixed, this is the only alternative I can see, not being a real kernel hacker myself. Let's just hope this stuff gets resolved upstream quickly.


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

Offline

#107 2012-11-01 10:38:31

pyknite
Member
Registered: 2010-03-03
Posts: 166

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

l3nkz, in fact it doesn't work anymore today oO after resume and reboot my machine still get high temperature sad

Offline

#108 2012-11-01 23:47:27

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 5,675

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@WonderWoofy,
Thanks again for your information. That's rather helpful. (Although I'm not actually using any of it at this point...)

WonderWoofy wrote:

As far as the snd_hda_intel, it is simply turning your sound "card" off when not in use.  The wort thing that can happen is that sometimes there is a delay for a split second when you start playing something with audio again.

I thought this was something laptop mode tools was meant to handle via the controller... but actually maybe that's generic. hmm...

Here s what I have for that... not sure if the file name is numbered right, but I figure 66 is a safe ways down the line, and it works.  It is a udev rule, so it is located in /etc/udev/rules.d/ :

% cat 66-sndpower.rules
#turn on intel sound card power saving feature (powertop recommended)
ACTION=="add", KERNEL=="snd_hda_intel", SUBSYSTEM=="module", TEST=="parameters/power_save", ATTR{parameters/power_save}="1"

Interesting. I compared this with the link you mentioned the other day on writing udev rules but either it didn't cover this or I couldn't see the connection. May I ask why you used a udev rule rather than setting an option for the module? I haven't actually rebooted yet to see if my change to the module options will do what I expect...

As far as the dirty_writeback_centisecs, I think it has something to do with the harddrive spin down and how often it spins back up to write.  I could be totally off on this, but I know it is what laptop mode is supposed to default to. 

For this one, I use /etc/sysctl.d with a file as follows:

%  cat vm-writeback.conf
## set /proc/sys/vm/dirty_writeback_centisecs to Powertop's recommended 1500
vm.dirty_writeback_centisecs=1500

"laptop mode is supposed to default to" in what sense? I mean, if it defaults to it, why would you need to set it explicitly?

Won't this method set that whether you are on battery or AC? I ask because unlike the audio one, I imagine there are significant disadvantages to this setting such as losing significantly more data in case of accident?


How To Ask Questions The Smart Way | Help Vampires

Arch Linux | x86_64 | GPT | EFI boot | grub2 | systemd | LVM2 on LUKS
Lenovo x121e | Intel(R) Core(TM) i3-2367M CPU @ 1.40GHz GenuineIntel | Intel Centrino Wireless-N 1000 | US keyboard with Euro | 320G 7200 RPM Seagate HDD

Offline

#109 2012-11-02 00:38:54

WonderWoofy
Member
From: Los Gatos, CA
Registered: 2012-05-19
Posts: 8,412

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@cfr,
Well I have never used laptop mode tools, so I just set all my stuff to save power and have not yet needed to have it otherwise.  So I am not sure if laptop mode really handles the audio thing or not.

As far as the sound rules... I probably should just set a kernel parameter.  I set this up a while ago, and when reviewing that particular one for the purposes of this thread, I didn't even really think about it.  That is probably something I should change though.  Mind you, the rule works, but probably is the incorrect way of doing it.

Honestly, I just thought i had read somewhere that this is what laptop mode tools defaulted to on the vm_writeback.  Laptop mode itself, you are right, you would not need to change it if it were the default.  I would imagine that yes, it will potentially lose much more data if you ran out of power somehow.  But I just put a second ssd in my machine in place of my old rotational drive.  So this actually makes no difference anymore (I don't think).  But even when I was using my old momentus xt, having all these things set yeilded nearly six hours of battery life, so I have only actually run out of power once.  Also, my data is not so very precious that I need to be worried about such things.  But now that I don't have a HDD in my machine anymore I actually get over six.  Of course those times are if I am not compiling or having anything heavily taxing the computer (like just basic internets, music, emails, etc.).  In your situation (with the overheating and all) I think that maybe it would in fact be better to set it so that you have less potential data loss.

Offline

#110 2012-11-02 00:57:00

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I'm about to install arch on my laptop with sandy bridge (toshiba portege r835 core i5-2410m). Is installing arch and using the 3.5 kernel a safe bet in terms of power usage and temperature? Are the boot parameters still relevant, or are these now included in more recent kernel updates?

Offline

#111 2012-11-02 08:07:23

l3nkz
Member
From: Dresden
Registered: 2011-02-01
Posts: 11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hi guys,

twistedcubic wrote:

I'm about to install arch on my laptop with sandy bridge (toshiba portege r835 core i5-2410m). Is installing arch and using the 3.5 kernel a safe bet in terms of power usage and temperature? Are the boot parameters still relevant, or are these now included in more recent kernel updates?

Hm, this really sucks. My laptop is running without any issues. I've used the patched linux 3.7rc2 for more than one week and resume/suspend has worked great.

hendry wrote:

yes, very nice tutorial there l3nkz. I didn't know about the mainline AUR package.

My machine is now 46C instead of almost 80C and it can be considered a laptop again smile

be nice to know when this patch is going to land in 3.7. I wonder why you didn't patch the linux package?

I like the linux-mainline package since it seems to have far less patches than the Arch linux one. Wonder what the need was to change the loglevel though.

thanks again

I didn't patched the linux package becasue I wanted to try the new kernel, but I can try the old one too. Or has anybody else tested it already?

eduedix wrote:

l3nkz, is yours X220,too?

No, i have a T420s.

Cheers

Offline

#112 2012-11-03 20:59:27

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

When I installed arch the other night, when it updated to 3.6.4 it would constantly run at the highest clock speed (even under a forced setting like powersave) and the temperature was geting extremely high (~79). As soon as I rolled back to 3.5.6 my temp stays at around 49-52 and the cpugovernor works as it should on the ondemand setting.

Offline

#113 2012-11-04 05:11:01

hendry
Member
From: Singapore
Registered: 2009-09-08
Posts: 127
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

This bug is nasty, I'm not convinced if it's properly fixed via the msleep patch or the latest 3.6.5.

Anyone know a good tool to simply log the temperature & the kernel revision somewhere? I'm using papertrailapp.com already for remote logging, I guess I would have to write some script or perhaps use sar.

Sidenote: Has anyone here running an X220 managed to do a BIOS update to 1.35? The BIOS update ISO is empty for me. http://support.lenovo.com/en_US/researc … MIGR-77280 https://twitter.com/kaihendry/status/264049382805094400

Offline

#114 2012-11-04 07:03:37

David Batson
Member
Registered: 2011-10-13
Posts: 640

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

hendry wrote:

Sidenote: Has anyone here running an X220 managed to do a BIOS update to 1.35? The BIOS update ISO is empty for me. http://support.lenovo.com/en_US/researc … MIGR-77280 https://twitter.com/kaihendry/status/264049382805094400

It's there, I just downloaded it.  Haven't done the update yet though...

http://download.lenovo.com/ibmdl/pub/pc … uj20uc.iso

Offline

#115 2012-11-04 07:40:03

hendry
Member
From: Singapore
Registered: 2009-09-08
Posts: 127
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

David Batson wrote:

It's there, I just downloaded it.  Haven't done the update yet though...

http://download.lenovo.com/ibmdl/pub/pc … uj20uc.iso


If you try mount the ISO, it's empty. Also I tried dd-ing to a USB stick and booting it, no-worky. sad

Offline

#116 2012-11-04 10:30:05

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

X220, Core(TM) i5-2450M CPU @ 2.50GHz with 3.6.5-1 kernel. The issue is still there.

But, when I upgraded to 3.6.5, the issue stopped for 2 or 3 reboots, the GPU was going to rc6 state, my CPU was also staying at 800MHz for some time. After this nasty bug came again, I have done 2-3 cold restarts, issue is still there. Also, I upgraded to the latest bios in a hope that something magical will happen, but no good.


EDIT: Ok, folks. I booted to lts 3.0, and then back to 3.6.5 and somehow everything is fine again. GPU is also going to RC6 state too (varying from 5-61%).

How is that possible, what changes b/w the restarts?

Last edited by shadyabhi (2012-11-04 10:37:03)

Offline

#117 2012-11-04 15:20:53

David Batson
Member
Registered: 2011-10-13
Posts: 640

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

hendry wrote:

If you try mount the ISO, it's empty. Also I tried dd-ing to a USB stick and booting it, no-worky. sad

Hmmm, I just burned a CD with the iso from the link and updated my BIOS to 1.35. wink

Offline

#118 2012-11-04 16:55:27

donniezazen
Member
From: Salt Lake City
Registered: 2011-06-24
Posts: 671
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I am running 3.7.0-1-mainline without patch and I don't have any problems. I didn't compiled or patched it but downloaded xz file from arch.miffe.org .

Offline

#119 2012-11-04 19:38:42

l3nkz
Member
From: Dresden
Registered: 2011-02-01
Posts: 11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

donniezazen wrote:

I am running 3.7.0-1-mainline without patch and I don't have any problems. I didn't compiled or patched it but downloaded xz file from arch.miffe.org .

How long have you used this kernel, because this just seems to be an 3.7rc3 kernel. Which is the same one as the one in the aur and according to
KaiSforza (see here)  this kernel doesn't fix the problem at all as well as the patch I posted.

After many syspend/resume cycles the problem returned for me too. Yesterday a had the old strange gpu activity problem again. -.-'
I really hope a fix will be published soon.

Offline

#120 2012-11-04 19:54:38

donniezazen
Member
From: Salt Lake City
Registered: 2011-06-24
Posts: 671
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@I3nkz I have been running it for almost two days. The whole kernel thing has been flaky. I have booted into a system that runs at high temperature all the time. I don't use any boot parameters as I couldn't find any conclusive evidence that they help. I am running pre-packaged vanilla kernel linux-mainline-3.7rc3-1-x86_64.pkg.tar.xz from miffe's website. I will let my system go into suspend a few times today and see if there are any problems. I will report back later today or tomorrow.

Offline

#121 2012-11-05 07:27:14

Rolinh
Member
From: Switzerland
Registered: 2011-05-07
Posts: 144
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I have been following this discussion for a while since I encounter the exact same problem with my Thinkpad T420 (Intel(R) Core(TM) i7-2620M CPU @ 2.70GHz).
I compiled kernel 3.6.5 as soon as it was out and installed it. I thought the problem was solved... but I was wrong (until today, temperature when idle was in the range 38°C-42°C).
Also, I removed all my kernel boot parameters (went from "i915.i915_enable_rc6=1 i915.i915_enable_fbc=1 i915.lvds_downclock=1" to nothing) and things seem a little better. However, behavior is not the same between two reboots... which I think is really weird and does not make sense at all.

Anyway, as apparently kernel 3.7 does not solve the issue yet, does anyone have kernel 3.5 64-bit around ?

Last edited by Rolinh (2012-11-05 07:29:13)

Offline

#122 2012-11-05 12:55:00

iDont
Member
Registered: 2009-10-22
Posts: 3

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Rolinh wrote:

Anyway, as apparently kernel 3.7 does not solve the issue yet, does anyone have kernel 3.5 64-bit around ?

You can find the last 3.5.x Arch kernel here: http://arm.konnichi.com/2012/10/15/core/os/x86_64/
(from https://wiki.archlinux.org/index.php/Do … kages#ARM)

Edit: by the way, another T420 user here. I'm also suffering from this annoying bug sad. Will stay on kernel 3.5.6 until it has been properly fixed.

Last edited by iDont (2012-11-05 12:57:44)

Offline

#123 2012-11-05 15:00:06

Rolinh
Member
From: Switzerland
Registered: 2011-05-07
Posts: 144
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@iDont: thanks, I had forgotten about arm since I had to actually use it.

3.6.6 is out but by reading the changelog, it does not seem that our issue got fixed.

Offline

#124 2012-11-06 01:50:06

Hspasta
Member
Registered: 2011-12-24
Posts: 187
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Both vanilla linux-mainline-rc4 and 3.6.6 yet again fail to fix this power issue...

Offline

#125 2012-11-06 02:16:32

hendry
Member
From: Singapore
Registered: 2009-09-08
Posts: 127
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

This is getting a little ridiculous. Isn't there a kernel bug # we can track? https://bugzilla.kernel.org/buglist.cgi … ndy+bridge

Offline

Board footer

Powered by FluxBB