You are not logged in.

#51 2012-10-21 20:16:06

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Well I guess my issue wasn't a kernel issue after all.

After updating the intel driver and xorg packages yesterday, my power consumption doesn't go ham after resuming from a suspend. I suppose though I'll leave this thread unsolved for those that actually have this problem.

Offline

#52 2012-10-21 23:32:05

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hspasta wrote:

Well I guess my issue wasn't a kernel issue after all.

After updating the intel driver and xorg packages yesterday, my power consumption doesn't go ham after resuming from a suspend. I suppose though I'll leave this thread unsolved for those that actually have this problem.

I'm going to take your word on this and update. If this stays consistent through a few sleep and restart cycles, I'd say we can mark it solved, and put a solution to update to the newest xorg packages.

EDIT:

I just updated. My power consumption is going haywire still. I think you may have gotten one of the lucky reboot cycles, Hspasta.

Last edited by KaiSforza (2012-10-21 23:41:08)


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

Offline

#53 2012-10-21 23:40:52

cfr
Member
From: Cymru
Registered: 2011-11-27
Posts: 5,662

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

WonderWoofy wrote:

You know all those i915 parameters you two are using?  Check out

# systool -m i915 -av

I set all those things at first, because I wasn't sure what the setting of -1 meant in those cases.  Apparently those things are defaulting to being on.  For instance, the i915_enable_rc6 can be set to 0, 1, 3, or 7.  1 will enable rc6, 3 will enable rc6p also, and 7 will enable rc6pp in addition to the other two.  But without setting it to 1 on my own, it actually sits in rc6p most of the time (which is a lower power state than rc6, and rc6pp is even lower). 

So I think that all the rc6 setting will actually set it to use ore power than the default.  Frame buffer compression appears to be on by default, as does semaphores.  The only one I think you are having a real effect with is the lvds_downclock setting.

I am not 100% on this.  But I noticed my Ivy Bridge machine actually uses less power when I don't set all that crap, though I have not tried the lvds_downclock parameter.

How do you figure out what i915 is using as defaults? I switched so I'm only forcing the downclocking bit explicitly but now can't tell what it is using as most of the parameters are "-1" for "use chipset's default" but how do I know what that is?

My feeling is, it is using more power now but feelings aren't a good guide and it could be for some other reason...


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

#54 2012-10-22 00:32:36

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Well I am unsure about figuring out some of the parameters, but I noticed in powertop that even when I didn't explicity enable rc6, it was still dropping it down the rc6p most of the time.  By default it doesn't seem to use rc6pp, but I actually get longer battery life when I don't use it.

Things like semaphores and lvds_downclock I am not sure.  I think I read on the intel graphics site that at least semaphores was enabled of compatible hardware by default.  But that was a while ago, so my memory is not fresh on that.

Offline

#55 2012-10-22 07:22:42

pyropirat
Member
From: Germany
Registered: 2012-10-22
Posts: 7

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hi,
yesterday i tried lts kernel to resolv the power consuming. Its working, but i dont think its a good solution. It feeles like on another hardware, realy slow. So i tried the linux-mainline kernel from aur. For me it is the best Solution for the Problem. So now i am back at 8W.

By the way: X220t with Intel i5@2,5Ghz with no bootkommands

Edit: Sorry i was wrong. speedstep is not working after suspend annymore. Same problem again.

Last edited by pyropirat (2012-10-22 10:38:55)

Offline

#56 2012-10-23 07:58:10

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I tried the linux-mainline kernel from the aur too and the problem is still there.
But clocking down the cpu works great again. My powerconsumption has fallen down to 6-7W when beeing in idle.

I've recognized a strange behaviour with the mainline kernel, maybe someone can verify it!
When I suspend for a long time (for example over the night) and then come back the scaling of the gpu works.
If i just suspend for a short period (about 5 mins) then the scaling doens't function. I tried it about ten times in the last two days.
Maybe I just had luck but if someone else can verify this behaviour this problem will become extremly strange.

BTW I have a Thinkpad T420s with Intel i5 and I use i915.i915_enable_rc6=-1 as bootparameter.

Offline

#57 2012-10-23 08:09:49

jrk
Member
From: Nämberch
Registered: 2012-10-16
Posts: 29

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

l3nkz wrote:

I've recognized a strange behaviour with the mainline kernel, maybe someone can verify it!
When I suspend for a long time (for example over the night) and then come back the scaling of the gpu works.
If i just suspend for a short period (about 5 mins) then the scaling doens't function. I tried it about ten times in the last two days.
Maybe I just had luck but if someone else can verify this behaviour this problem will become extremly strange.

That's quite interesting. Maybe it's something about the gpu cooling down to room temperature which flips a certain bit (sloppy spoken..).
After all this is hardware. I know why I usually try to keep my fingers of that stuff.. smile
Could be completely unrelated though because hardware.

Maybe you could run a little test series like so:

Suspend
Wait {5,10,15,20,25,30, as long/often you want too}
Resume
Check temperature (write a script which dumps the gpu temps right after resuming)
Check frequency scaling

If you can do this, it would be great if you could post your results also to the kernel bug report.

Thanks!

Offline

#58 2012-10-23 08:27:20

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

jrk wrote:

Maybe you could run a little test series like so:

Suspend
Wait {5,10,15,20,25,30, as long/often you want too}
Resume
Check temperature (write a script which dumps the gpu temps right after resuming)
Check frequency scaling

If you can do this, it would be great if you could post your results also to the kernel bug report.

Thanks!

That's an interesting idea. I'll do it as soon as I will find some time to do it. At the moment life's beeing quite stressfull, but I'll keep it in mind!

Cheers!

Offline

#59 2012-10-23 16:32:00

flockyrocky
Member
From: China
Registered: 2012-09-15
Posts: 75

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

wow, it's good to know I'm not the only one who's experiencing this issue.
My T420s has exactly the same issue with you guys, the only difference is that it only happens to me sometimes after resuming from suspend. Seems like I had to downgrade to earlier kernel.

Offline

#60 2012-10-24 10:41:23

Hexal
Member
Registered: 2012-10-24
Posts: 1

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Unfortunately with 3.6.3 it's even worse.

Offline

#61 2012-10-24 11:04:27

curson
Member
From: London
Registered: 2010-12-03
Posts: 32
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

On my Dell XPS 14z, I'm experiencing an increase in average idle temperature of around 10~15°C on my usual straight after boot (with relative wild increase in fan spin noise/speed).

The problem appeared today as I upgraded from 3.6.2 to 3.6.3 kernel. Both with old and current version of xorg-server everything is normal (fine, with idle temperature around the usual 48°C) with kernel 3.6.2-1-ARCH but goes sort of crazy with 3.6.3-1. I haven't tested it thoroughly as it kind of scared me and I opted for a quick downgrade (I'm on the road, and I need my computer atm, no proper time to look into it). Gladly that fixed it for me.

Just as additional info: the system is configured with nvidia-bumblebee, so I'm running with i915 as standard, activating the Nvidia GPU just when needed through bbswitch and 'optirun' and every package is up-to-date (with the obvious exception of the kernel).

Offline

#62 2012-10-24 13:30:44

serialx2
Member
From: Sao Paulo, Brazil
Registered: 2012-04-21
Posts: 1
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Thinkpad X1 is also affected (Intel i5-2520M 2.5GHz)

After returning from sleep mode it have the highest power consumption, but during normal operation cpu scaling does not work.

Offline

#63 2012-10-24 13:53:42

Anthony25
Member
Registered: 2012-10-24
Posts: 43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hello,
I have the same problem with 3.6.2 and 3.6.3 ...

It seems that the ondemand governor doesn't work, so it switch directly from the lowest frequency (for me 800Mhz) to the highest frequency (2,3Ghz), and not by levels like before.

Offline

#64 2012-10-24 14:12:27

glock24
Member
Registered: 2011-01-28
Posts: 6

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I have a Lenovo y470p with hybrid graphics (i7-2670QM and HD7690M), and for me the power usage seems a bit high at ~17W idling. I'm using systemd-gaswitcheroo-units 1.1-1 from AUR to disable the discrete GPU. I'll update to kernel 3.6.3 and check for changes. My kernel boot parameters are:

i915.i915_enable_rc6=7 i915.i915_enable_fbc=1 i915.lvds_downclock=1 i915.semaphores=1 i915.modeset=1

Not using those parameters result in ~1-2W increase at idle.

EDIT: Same behaviour with kernel 3.6.3 as before.

Last edited by glock24 (2012-10-24 14:18:22)

Offline

#65 2012-10-24 14:31:46

Anthony25
Member
Registered: 2012-10-24
Posts: 43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

If you want to save your battery, switch on the conservative governor, it's slower than the ondemand but at least it works.

Offline

#66 2012-10-24 15:06:08

Gusar
Member
Registered: 2009-08-25
Posts: 3,450

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Anthony25 wrote:

If you want to save your battery, switch on the conservative governor, it's slower than the ondemand but at least it works.

No, the conservative governor does *not* have the effect you think it has.

https://bbs.archlinux.org/viewtopic.php?id=151263

Online

#67 2012-10-24 18:27:20

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

3.6.3-1 fixed it for me.

Lenovo Thinkpad X220 here. Now, it stays at arounf 43-50 C during normal usage.

Last edited by shadyabhi (2012-10-24 18:27:37)

Offline

#68 2012-10-24 19:01:37

Anthony25
Member
Registered: 2012-10-24
Posts: 43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Gusar wrote:

No, the conservative governor does *not* have the effect you think it has.

https://bbs.archlinux.org/viewtopic.php?id=151263

I didn't know that it was made for old CPU, but I know it's less efficient than the ondemand governor (that's why I never switch on conservative).

But now with linux 3.6.3 it's the only governor who changes the frequency by levels and not in a binary way (freq mini or freq maxi), even if, I agree, it's not smooth with it, but still better than the powersave governor...

Offline

#69 2012-10-24 19:11:20

vheissu
Member
From: Vienna
Registered: 2012-10-24
Posts: 3

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

shadyabhi wrote:

3.6.3-1 fixed it for me.

Lenovo Thinkpad X220 here. Now, it stays at arounf 43-50 C during normal usage.

sry, can’t confirm that, it’s still not working with my X220 and ondemand governor after suspend.

Offline

#70 2012-10-24 19:56:44

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Strangely, my system temperature is 70 C on linux-lts and 85 C on 3.6.3-1

Last edited by donniezazen (2012-10-25 01:12:56)

Offline

#71 2012-10-24 23:06:15

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

3.6.3-1 seems to be working for me on this boot. I'll do some suspend/resume/reboot cycles to test and see if it holds, though.

donniezazen wrote:

Strongly, my system tempreature is 70C on linux-lts and 85C on 3.6.3-1

donniezazen: did you boot into that power usage, or was that after a suspend/resume?

Last edited by KaiSforza (2012-10-24 23:06:28)


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

Offline

#72 2012-10-24 23:20:46

Mindstormscreator
Member
Registered: 2012-07-01
Posts: 165

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza wrote:

I'll do some suspend/resume/reboot cycles to test and see if it holds

I would be cautious about doing this... http://www.h-online.com/open/news/item/ … 36110.html

And also to anybody else trying to test their temperatures by frequent reboots.

Offline

#73 2012-10-24 23:24:38

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Mindstormscreator wrote:
KaiSforza wrote:

I'll do some suspend/resume/reboot cycles to test and see if it holds

I would be cautious about doing this... http://www.h-online.com/open/news/item/ … 36110.html

And also to anybody else trying to test their temperatures by frequent reboots.

Well, looks like I've finally got a reason to upgrade to btrfs. Butterface, here I come!


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

Offline

#74 2012-10-24 23:35:27

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza wrote:

3.6.3-1 seems to be working for me on this boot. I'll do some suspend/resume/reboot cycles to test and see if it holds, though.

donniezazen wrote:

Strongly, my system tempreature is 70C on linux-lts and 85C on 3.6.3-1

donniezazen: did you boot into that power usage, or was that after a suspend/resume?

I booted to high system temperature. I am now using 3.7.0-1-mainline and it seems to work fine.

Offline

#75 2012-10-25 00:35:35

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

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

donniezazen wrote:
KaiSforza wrote:

3.6.3-1 seems to be working for me on this boot. I'll do some suspend/resume/reboot cycles to test and see if it holds, though.

donniezazen wrote:

Strongly, my system tempreature is 70C on linux-lts and 85C on 3.6.3-1

donniezazen: did you boot into that power usage, or was that after a suspend/resume?

I booted to high system temperature. I am now using 3.7.0-1-mainline and it seems to work fine.

No patches to the mainline I take it?

EDIT: Also, suspended, resumed, checked powertop and saw my cores at max frequency again. Gonnay try it with 3.7.0-1.

EDIT OF THE EDIT: Trying out 3.7.0-1-mainline. So far, through a suspend resume cycle, it's maintained a consistent power use and temp.

Last edited by KaiSforza (2012-10-25 02:23:49)


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

Offline

Board footer

Powered by FluxBB