Strange FID & VID - Values

If you have trouble with PHC you can ask and hope for help here.
Spindoctor
Posts: 13
Joined: Thu 20. Nov 2008, 03:41

Strange FID & VID - Values

Post by Spindoctor » Thu 20. Nov 2008, 03:57

Hello!

I'm just about to undervolt my T9400 CPU.

But using phc some things seem strange to me.

First of all the FID-values and VID-values.

Code: Select all

# cat /sys/devices/system/cpu/cpu0/cpufreq/phc_controls
gives me the values

Code: Select all

10:43 73:36 6:23 134:15
If I understand this right, my possible multiplier settings are 10, 73, 6 and 134. These values would be a bit strange ;-)

Next, the VID-values are also somehow strange as the highest FID has the lowest VID (134:15) and the second-lowest FID has the highest VID (10:43).

In phctool this all makes more sense as the values are
2534Mhz:VID=43, 2533Mhz:VID=36, 1600Mhz:VID=23 and 800Mhz:VID=15.

However what sense does it make to have 2533Mhz and 2534Mhz as possible values?
And why do they have such a big voltage difference?

Also I discovered that I can go really low with the VID-settings.
1600Mhz and 800Mhz both take VID=0.
I'm just about to check out 2533Mhz (and 2534Mhz) but they seem to take something arround VID=20. Is this difference normal?

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

Re: Strange FID & VID - Values

Post by the-fallen » Thu 20. Nov 2008, 07:26

Hello,

yes this problem is known but we do not have a solution yet.

Would be easier to have some access to such a CPU.


I try to write a small tool to dump the needed part from MSR's. Maybe you can use it and give me the output.

Spindoctor
Posts: 13
Joined: Thu 20. Nov 2008, 03:41

Re: Strange FID & VID - Values

Post by Spindoctor » Thu 20. Nov 2008, 10:01

if it doesn't harm my CPU (or any other part) in any way, I could do that.

But: which problem is known?

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

Re: Strange FID & VID - Values

Post by the-fallen » Thu 20. Nov 2008, 10:16

The problem that the FIDs look strange is known. Maybe it is caused by a shifted bit or a FID wider than 8Bit. I am going to extend the read_msr tool to dump the interesting part from MSR.


I do not think that this could harm your CPU because the VIDs look normal and at least the CPU will protect itself but I can not say that I am sure so it is your risk and you need to consider that fact.

Spindoctor
Posts: 13
Joined: Thu 20. Nov 2008, 03:41

Re: Strange FID & VID - Values

Post by Spindoctor » Thu 20. Nov 2008, 13:38

as I don't know your real name, I couldn't call you to account anyway... ;-)

so if you really think it is safe to try it, I would do it. tell me what to do.

But anyway, can you tell me why I have 2533 Mhz and 2534 Mhz steps and why their default voltage has such a big difference?

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

Re: Strange FID & VID - Values

Post by the-fallen » Thu 20. Nov 2008, 17:23

Please ignore the frequency values. They are not correct because of the strange FIDs.
The VIDs seem to be okay so please only use the "phc_vids" - Interface and not the "phc_controls" - Interface.

Can you give me the default frequencies for your CPU please?

You are using PHCTool, right?

Spindoctor
Posts: 13
Joined: Thu 20. Nov 2008, 03:41

Re: Strange FID & VID - Values

Post by Spindoctor » Thu 20. Nov 2008, 20:34

the-fallen wrote:Can you give me the default frequencies for your CPU please?
of course but how do I find them?
As I already mentioned, phc_fids gives 10 73 6 134 (which seems to be incorrect).
phctool tells me the following about my CPU:
Name | Intel(R) Core(TM)2 Duo CPU T9400 @ 2.53 GHz
CPU Model | Vendor:GenuineIntel, Family:6, Model:23, Stepping:6
Frequencies | 2534Mhz 2533Mhz 1600Mhz 800Mhz
the-fallen wrote:You are using PHCTool, right?
beside other tools, yes.

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

Re: Strange FID & VID - Values

Post by the-fallen » Thu 20. Nov 2008, 20:55

Just enter the following command into a console :

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies

PHCTool calculates the frequencies from given FIDs and a formula that depends on the type of CPU. This can be wrong for newer CPU Types of if - like in your case - the FIDs are somehow incorrect.

But PHCTool just displays the data - nothing more.

How ever, thank you for your support :)

Spindoctor
Posts: 13
Joined: Thu 20. Nov 2008, 03:41

Re: Strange FID & VID - Values

Post by Spindoctor » Thu 20. Nov 2008, 21:03

you are welcome.
And thank you for your support :)

the output of scaling_available_frequencies is not really spectacular:
2534000 2533000 160000 800000

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

Re: Strange FID & VID - Values

Post by the-fallen » Sat 22. Nov 2008, 09:02

Thanks for the data.

But they are quite spectacular.

I thought the patch somehow reads the FIDs wrong in some cases (those unnatural high FIDs you posted) but now it seems that this problem is somewhere in the kernel, not within phc because those frequencies you posted look unnatural.

I will investigate that a bit more.

Post Reply