all voltages on zero but still running

If you have trouble with PHC you can ask and hope for help here.
sonium
Posts: 1
Joined: Thu 21. May 2009, 12:35

all voltages on zero but still running

Post by sonium » Thu 21. May 2009, 12:41

Hi,
it seems that somehow the undervolting process has no effect.
I'm using the precompiled kernel from https://launchpad.net/~linux-phc/+archive/ppa on ubuntu-9.04 amd64. Additionaly there is no acpi_cpufreq
or speedstep_centrino module listed, but makes sense as it is probably in the kernel. I have tried undervolting both cores to zero but still no effekt. Even tried disabling one core.

I'm running on a Core 2 Duo.

So, any ideas what might be wrong?

the-fallen
Administrator
Posts: 346
Joined: Wed 9. Jul 2008, 19:57

Re: all voltages on zero but still running

Post by the-fallen » Thu 21. May 2009, 20:38

You can check what values are assigned to a CPU using the tool "read_msr" (downloadable here in the forum).

This will help a lot to figure out where the problem could be.

evil
Posts: 27
Joined: Sun 20. Sep 2009, 17:46

Re: all voltages on zero but still running

Post by evil » Sun 20. Sep 2009, 17:56

Hi, I think i have the same problem, but I'll report completely again.
Intel T9550

I noticed the weird 134:19.
Anyway I tried to stress both cpus to find the minimum VID but at first round it went down to -1 and stopped.
After a few seconds it froze and reboot by itself, and I cannot explain that.
How do you suggest to proceed to understand why it rebooted and which values to use anyway?
evil@hell:~/Download/read_msr 0.2pre$ uname -a
Linux hell 2.6.28-15-generic #50~undervolt2-Ubuntu SMP Fri Aug 21 10:34:50 UTC 2009 i686 GNU/Linux

evil@hell:~/Download/read_msr 0.2pre$ cat /sys/devices/system/cpu/cpu0/cpufreq/phc_controls
74:43 10:37 8:30 6:23 134:19

evil@hell:~/Download/read_msr 0.2pre$ sudo ./read_msr.py --cpuinfo
MSRTOOL V0.2pre started...

trying to load msr module

cpu0
highest frequency: 2668 MHz
lowest frequency: 800 MHz
PHC Version: 0.3.2-1
cpu1
highest frequency: 2668 MHz
lowest frequency: 800 MHz
PHC Version: 0.3.2-1

evil@hell:~/Download/read_msr 0.2pre$ sudo ./read_msr.py --availfids
MSRTOOL V0.2pre started...

cpu0
Your PHC Version is < 0.4 - the available_fids interface is not supported
cpu1
Your PHC Version is < 0.4 - the available_fids interface is not supported
(but this is what the ppa kernel was shipped with...)

evil@hell:~/Download/read_msr 0.2pre$ sudo ./read_msr.py --readmsr
MSRTOOL V0.2pre started...

Displaying data for CPU cpu0
/---------------------------------------
| Current VID: 19
| Current FID[DID]: 6[1]
| Target VID: 19
| Target FID[DID]: 6[1]
| Highest VID: 43
| Highest FID[DID]: 10[0]
| Lowest VID: 23
| Lowest FID[DID]: 6[0]
\---------------------------------------



Displaying data for CPU cpu1
/---------------------------------------
| Current VID: 19
| Current FID[DID]: 6[1]
| Target VID: 19
| Target FID[DID]: 6[1]
| Highest VID: 37
| Highest FID[DID]: 10[0]
| Lowest VID: 23
| Lowest FID[DID]: 6[0]
\---------------------------------------
more info needed?

evil
Posts: 27
Joined: Sun 20. Sep 2009, 17:46

Re: all voltages on zero but still running

Post by evil » Mon 28. Sep 2009, 20:35

nothing?

the-fallen
Administrator
Posts: 346
Joined: Wed 9. Jul 2008, 19:57

Re: all voltages on zero but still running

Post by the-fallen » Tue 29. Sep 2009, 16:48

Heya,

I am sorry for answearing that late.

First, you are not using phc 0.4 because there is not a phc-patch v 0.4 for Intel at the moment. The PPA is shipped with intel 0.3.2 and amd 0.4 but you have an intel CPU.

That your System crashed by a VID lower than 0 ( you said '-1' ?) does not surprise me. But good to know that we need to change the sanity checks there :)

You may try to manually set frequencies and VIDs without using a script (you used a script, don't you? Sounded like that to me).
You also should try that at a frequency that is not the lowest one because the VID for the lowest frequency can not be lowered. This one is the absolute minimum on a CPU while running and seems to be hard-capped in the CPU.

evil
Posts: 27
Joined: Sun 20. Sep 2009, 17:46

Re: all voltages on zero but still running

Post by evil » Tue 29. Sep 2009, 18:45

well, can't i undervolt with phc < 0.4?

and yes, i used the automatic test script from ubuntuforums.
what surprised me about the crash is that it reached -1 and "after a while" it crashed.

the test started with the highest freq, and for that one it reached -1.
indeed in my previous laptop i remember i set to 1 (because of the hard-cap) all freqs, but not the highest, which had higher VID.

so i think that manually i would do the same the script did, going down to -1 with the highest freq, and crash after some time, making no sense :)

the-fallen
Administrator
Posts: 346
Joined: Wed 9. Jul 2008, 19:57

Re: all voltages on zero but still running

Post by the-fallen » Wed 30. Sep 2009, 17:57

Sure, you can Untervolt with a patch < 0.4. Its just that read_msr 0.2 is ready for features that will come with 0.4 and told you that you missed those features. Nevertheless you should be able to use phc 0.3 with your CPU.

That it crashed after a while may be caused by the circumstance, that the value "-1" is used when the CPU reaches the frequency you set the voltage for.

For example if you set -1 for the highest frequency the value will be evaluated if the System got the information to switch to that frequency. If you already had been to the highst frequency, the new value will also only be used after you leave and re-enter this frequency. That may be the reason for "after a while".

You shall not go down to -1 manually. Just decrease the VID by 1 manually and (re-)enter the frequency. Then check it with read_msr. To be at the save side, reduce the VID for each core of the CPU by 1, enter the frequency and check then (shouldn't be necessary but who knows...)

I hope we can solve your problem :)

evil
Posts: 27
Joined: Sun 20. Sep 2009, 17:46

Re: all voltages on zero but still running

Post by evil » Wed 30. Sep 2009, 18:40

i understand you point but i guess i need to explain better.
the script started from the highest frequency (the first in the phc_default, or whatever is the name :) ), with both CPUs under 100% stress.
it reached a -1 VID and said that it is a wrong state.
good, it means that under 100% stress on both CPUs, the machine can bear the minimum voltage without crashing.
but, after a while, it crashed, so the question is:
why under big stress the machine didn't crash, but it did after a while with normal operational load?

the-fallen
Administrator
Posts: 346
Joined: Wed 9. Jul 2008, 19:57

Re: all voltages on zero but still running

Post by the-fallen » Wed 30. Sep 2009, 21:46

Thats why I wanted you to manually set a new VID and check it with read_msr so we can se if the CPU really uses a lower VID or if the system ignores the VIDs and just crashes because of the faulty "-1"

evil
Posts: 27
Joined: Sun 20. Sep 2009, 17:46

Re: all voltages on zero but still running

Post by evil » Wed 30. Sep 2009, 21:52

ok, could you point me to a brief howto please? (so far i used only the script and it worked)
i'll let you know what happens then.
thanks

Post Reply