user undervolting

just-linux · n00b Joined: 23 Mar 2005 Posts: 17

For my Intel Pentium M Dothan (2000MHz), undervolting does quite a big difference. At full frequency, It drops the temperature by approximately 20°C

I haven't had a look at the battery life yet. However, as soon as I have to travel again, I'll check this out

@ info2
Thanks for for the hints

But still, I can't set my VIDs lower than 18 :S And I'm pretty sure that my Bios is not broken...

shladnik · n00b Joined: 13 Feb 2007 Posts: 13

Hi all,

I was using patch 0.2.9 happily, but recently i decided to upgrade kernel (linux-2.6.20-suspend2-r6 -> linux-2.6.21-suspend2-r7), so i downloed newer patch also.
I see the files:

info2 · n00b Joined: 21 May 2007 Posts: 11

Hello

shladnik:
There are some users that have this problem.
The reason is that acpi-cpufreq decides if your system is either "SYSTEM_IO_CAPABLE" or "SYSTEM_INTEL_MSR_CAPABLE". It seems that especially on older hardware acpi-cpufreq switches to the first while linux-phc need the second state.

We are currently reworking the speedstep-centrino patches. They will be released with linux-phc-0.3.1.
They will have the same sysfs interface like acpi-cpifreq so PHCtool and other userspace programs can be used with it.
Until these patches are completed you should use the 0.2x release.

If you like to test the new patches: we need you

In some few weeks the first test release will be ready.

just-linux:
Very strange.
Could you compile your kernel with cpufreq.debug enabled and send me the output of dmesg with all cpufreq messages in it (after trying to set a vid below 18, etc to have everything in it)?
Maybe i can see whats wrong.
Did it worked with the old patches and speedstep-centrino? Maybe you're also an candidate for 0.3.1.

Schmolch:
You are the first one with an lv-cpu.
It would be quite interesting to see your changes (standard-vids and lower vids, temperature difference and battery time difference, if any).
If you like you can post them here of meet me in the IRC channel of linux-phc.

Like i already wrote, we'll need people to test the coming speedstep-centrino patches once they are ready.

just-linux · n00b Joined: 23 Mar 2005 Posts: 17

@info2

Recently, I tried again to set a VID lower than 18 and it WORKED! I don't really know why, I haven't changed anything...

info2 · n00b Joined: 21 May 2007 Posts: 11

Great news!
Maybe it's because the sun is in line with Venus and there was a mice running over a hill in Afrika.

info2 · n00b Joined: 21 May 2007 Posts: 11

Tester needed!

I'm proud to tell you that linux-phc-0.3.1 is nearby complete.
We just wish to test it on some different systems / processors to be sure it'll work for everybody.
Whats new?
* The speedstep-centrino patches has been reworked.
They now use the new interface acpi-cpufreq already use and is compatible with PHCtool.
* Also this patch doesn't just reject values out of range of 0.700 - 1.600Volts, it is rejecting all values above the default voltage of each step.
Accidently overvolting is now not even possible by a microvolt. Acpi-cpufreq patches already own this check.

I'll upload the developement release as soon as possible in svn.
People that want to test it now can ask for it on the irc-channel.

TheKaneB · n00b Joined: 15 Sep 2007 Posts: 2

Hi!

I'm the new entry who reworked the 0.2.x patch into the 0.3.1 patch... I'm here to get any feedback directly from you.

@info2: do we know each other? ^_^
_________________
Software Failure: Guru Meditation

Schmolch · l33t Joined: 16 Jun 2002 Posts: 746 Location: Germany

shladnik · n00b Joined: 13 Feb 2007 Posts: 13

I can report that new patch works for me.

I used the patch from svn:
linux-phc/trunk/devel/kernel-patch/speedstep-centrino/linux-phc-0.3.1-kernel-vanilla-2.6.21.7_sc.patch

I am using kernel linux-2.6.21-suspend2-r7 (gentoo system of course) with following processor:

info2 · n00b Joined: 21 May 2007 Posts: 11

Thanks a lot, shladnik.
Using this new patch and the new interface you should also be able to use PHCtool.

shladnik · n00b Joined: 13 Feb 2007 Posts: 13

Well, i can't try it right now, my system crashed... :cry:

don't know exaclty what happened. Probably something about hibernate or switching between X and console. Is there any possibility that patch would cause that? However, filesystem is corrupted (reiserfs), --rebuild-tree didn't help, so i will have to reinstall gentoo. Well, there are some files still readable. Pray for my /home and /etc.

Stefan

info2 · n00b Joined: 21 May 2007 Posts: 11

shladnik · n00b Joined: 13 Feb 2007 Posts: 13

It was suspend2 probably or simply write cash of disk wasn't flushed.
For sure it wasn't undervolting too deep, since i am using this configuration for over a year without problems. And even if cpu would crash, there is probably no chance that this would affect file system.

superwutze · Posted: Tue Sep 25, 2007 8:59 pm Post subject:

well, i have to admit that i didn't read the whole thread, so maybe i missed it, but in a quick scan i haven't read anyone mentioning the x86_64 platform.

i have a hp 6510b with an intel t7700 in a pure 64bit environment.
i patched the kernel (gentoo-2.6.22-r6) successfully but there were no options in the config menu. also i use the acpi-scaling driver right now. i now, the 64bit centrinos are pretty young, but will there be support for them?
on my last notebooks cpupw (pentium-m, 1g3) and the powernow-k8 patch (turion tl-56) worked very good and i just dislike alltime-running-fans :-)

so far i don't even know the available frequency/voltage pairs. cpupw just fails and on vista64 rmclock doesn't work either.
_________________
bill who? micro what?

info2 · n00b Joined: 21 May 2007 Posts: 11

Hello superwutze.

We never tried to apply those patches to a pure 64Bit system.
We would need developers and testers with 64Bit machines to maintain 64Bit patches.
Technically it >>could<< work.

TheKaneB · n00b Joined: 15 Sep 2007 Posts: 2

Hi superwutze,

as info2 said, to write a driver for your cpu we need a developer who owns that hardware. However, we are working to improve the old powernow-k8 patch for amd64 platforms, in order to be able to control every platform (centrino, core, amd64) with a generic platform-independent interface, so we can use the same tool for every machine. If anyone with good coding skills and the right hardware wants to help, please contact me or info2.

bye!
_________________
Software Failure: Guru Meditation

bushwakko · Guru Joined: 25 Mar 2003 Posts: 495

I have a macbook pro 2.33ghz (not santa rosa) and I have undervolted in osx like this: http://generation.no/~wakko/undervolt.png and using phctool to find the same settings in linux and applying them made my laptop lock up pretty fast, so I'm thinking something isn't quite right. anyone can confirm this?

edit: I'm running 64-bit ubuntu at the moment, with the kamikaze patchset. I think they might use an old patch (2.6.22-pre1) but I'm not sure.
_________________
Macbook Pro 15"
Core 2 Duo 2,33ghz
2gb Ram
ATI Mobility x1600 256mb

tnt · Veteran Joined: 27 Feb 2004 Posts: 1227

hm... many core2duo users run 64bit gentoos. it would be very nice if there was some 64bit version of PHC...
_________________
gentoo user

tnt · Veteran Joined: 27 Feb 2004 Posts: 1227

btw, I've just found this great site:

http://www.lesswatts.org/

and had to share it with you... :lol:

_________________
gentoo user

vovin · n00b Joined: 07 May 2006 Posts: 16

I tried to change voltage using patch 0.3.1 on
Linux localhost 2.6.23.9-phc #1 SMP Sat Dec 8 01:00:39 CET 2007 x86_64 Intel(R) Core(TM)2 Duo CPU E6750 @ 2.66GHz GenuineIntel GNU/Linux

It seems to detect CPU:
affected_cpus
0
cpuinfo_cur_freq
2664000
cpuinfo_max_freq
2664000
cpuinfo_min_freq
1998000
phc_controls
8:42 6:29
phc_default_controls
8:42 6:29
phc_default_vids
42 29
phc_fids
8 6
phc_version
0.3.1
phc_vids
42 29
scaling_available_frequencies
2664000 1998000
scaling_available_governors
ondemand userspace
scaling_cur_freq
2664000
scaling_driver
acpi-cpufreq
scaling_governor
userspace
scaling_max_freq
2664000
scaling_min_freq
1998000
scaling_setspeed
2664000

If I change phc_vids, there is no change in voltage and idle temperature according to gkrellm2.
If I change voltage in BIOS, I can see lower voltage and idle temperature in gkrellm2.

dixon · n00b Joined: 05 Jan 2008 Posts: 1

I've tried out the latest linux-phc 0.3.1 running ubuntu gutsy gibbon(CPU T5470). Everything was working fine except for one thing. I cannot decrease the voltage of the lowest frequency and this frequency is the mostly used

If I change it to any lower value the phctool analysis tab gives me ->
target vid - 14
target vcc - 0.8875V
reached vid - 15
reached vcc - 0.9000V
Though lowering the voltages of other(higher) frequencies is working just fine :\

SteveBallmersChair · Posted: Tue Jan 08, 2008 4:14 am Post subject:

gesho · n00b Joined: 11 Jan 2008 Posts: 4

hi folks, thanks for your work.

now i'm runnning 2.6.23-r5, the phc-0.3.0 tarball had nothing for this one (r3 was the last). anything planned?

marcogrego · n00b Joined: 10 Mar 2008 Posts: 5

hi everybody,
I am a new entry, and I happly tried phctool with GUI. I am able to set values under the "voltage" tab, but unfortunately I got a "n/a" for all the columns for the "Analysis" tab. Which makes me think my system is not correctly recognized (?).
My aim is just checking whether the undervolting I am doing is applied.
I also did patch the acpi-cpufreq kernel module and recompiled a custom kernel excluding the deprecated speedstep-centrino. (my kernel: 2.6.22)
Any advice?

Thank you in advance!

--
Marco

SteveBallmersChair · Posted: Tue Mar 11, 2008 1:10 pm Post subject: Re: undervolting through phctool