Success T7200 2GHz D820 Laptop,Ubuntu 9.10 HALFPOWER

You can write down your success stories here (temperature decreasements, power savings, etc).
Post Reply
jm.mico
Posts: 2
Joined: Sun 8. Nov 2009, 12:03

Success T7200 2GHz D820 Laptop,Ubuntu 9.10 HALFPOWER

Post by jm.mico » Sun 8. Nov 2009, 12:24

Hello people who make my life nicer :)

I am a bit of a freak for noise and temperature, so after I had Jaunty working properly and tuned with powertop and undervolted, I had to redo some of the work after upgrading to karmic.

Here is a little table that I made some time ago as estimation for the power savings when using my new VID table.

Code: Select all

                B E F O R E  O P T I M I Z A T I O N           A F T E R  P H C
Freq MHz       VID    Voltage(mV)         Power(mW/ß)              VID    Voltage(mV)        Power(mW/ß)
1000            19         1004                 1008                 1       716                      513
1333            27         1132                 1708                 1       716                      513
1666            35         1260                 2645                20       1020                    1733
2000            43         1388                 3853                30       1180                    2785
So as you can see, I have changed on the lower freqency (which is where the CPU is 95% of the time) power consumption to almost HALF!

In order to calculate the power savings I have assumed that the power consumption of the cores is approximately increasing linearly with the frequency and quadratic with Voltage. Those numbers are far from being a real physic unit but give an idea about the factor of reduction.

COOL!!!

And rock-solid stable!

Thanks for the tool ;)


Edited by the-fallen: Made the Table better readable.

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

Re: Success T7200 2GHz D820 Laptop,Ubuntu 9.10 HALFPOWER

Post by the-fallen » Sun 8. Nov 2009, 13:31

That are really nice results and an intersting table.

I did not understand why you are able to set VIDs down to "1".

Have you checked the applied FID/VID with read_msr?
Or do you have some special hardware?

normally the Cores have a hard-lock on the lowest VID
so one is normally not able to use depper values.

Would be interesting if that is really applied to your CPU.

jm.mico
Posts: 2
Joined: Sun 8. Nov 2009, 12:03

Re: Success T7200 2GHz D820 Laptop,Ubuntu 9.10 HALFPOWER

Post by jm.mico » Sun 8. Nov 2009, 17:56

Thanks for editing the table. I did not click on the preview button and when writting it looked OK.

I think I hoped more that I actually had...using the read_msr pyton script shows that, as you pointed, the hw may be locking some undervoltage settings. Here are the reading for my settings (1Ghz,1.33,1.66,2.0)

1GHz

Code: Select all

[cpu0] [CURRENT] FID:6 HID:0 DID:0 VID:19 
[cpu0] [TARGET]  FID:6 HID:0 DID:0 VID:1 
[cpu0] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu0] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu0] [SLFM]    FID:6 VID:19 
[cpu0] [IDA]     FID:12 VID:43 
[cpu0] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1

[cpu1] [CURRENT] FID:6 HID:0 DID:0 VID:19 
[cpu1] [TARGET]  FID:6 HID:0 DID:0 VID:1 
[cpu1] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu1] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu1] [SLFM]    FID:6 VID:19 
[cpu1] [IDA]     FID:12 VID:43 
[cpu1] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1
1.33GHz
[cpu0] [CURRENT] FID:8 HID:0 DID:0 VID:19
[cpu0] [TARGET] FID:8 HID:0 DID:0 VID:1
[cpu0] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu0] [LOWEST] FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu0] [SLFM] FID:6 VID:19
[cpu0] [IDA] FID:12 VID:43
[cpu0] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1

[cpu1] [CURRENT] FID:8 HID:0 DID:0 VID:19
[cpu1] [TARGET] FID:6 HID:0 DID:0 VID:1
[cpu1] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu1] [LOWEST] FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu1] [SLFM] FID:6 VID:19
[cpu1] [IDA] FID:12 VID:43
[cpu1] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1
1.66GHz

Code: Select all

[cpu0] [CURRENT] FID:10 HID:0 DID:0 VID:20 
[cpu0] [TARGET]  FID:10 HID:0 DID:0 VID:20 
[cpu0] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu0] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu0] [SLFM]    FID:6 VID:19 
[cpu0] [IDA]     FID:12 VID:43 
[cpu0] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1

[cpu1] [CURRENT] FID:10 HID:0 DID:0 VID:20 
[cpu1] [TARGET]  FID:6 HID:0 DID:0 VID:1 
[cpu1] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu1] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu1] [SLFM]    FID:6 VID:19 
[cpu1] [IDA]     FID:12 VID:43 
[cpu1] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1
And 2GHz

Code: Select all

[cpu0] [CURRENT] FID:12 HID:0 DID:0 VID:30 
[cpu0] [TARGET]  FID:12 HID:0 DID:0 VID:30 
[cpu0] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu0] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu0] [SLFM]    FID:6 VID:19 
[cpu0] [IDA]     FID:12 VID:43 
[cpu0] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1

[cpu1] [CURRENT] FID:12 HID:0 DID:0 VID:30 
[cpu1] [TARGET]  FID:6 HID:0 DID:0 VID:1 
[cpu1] [HIGHEST] FID:12 (HID:0 DID:0) VID:43 (not sure if they exist here)
[cpu1] [LOWEST]  FID:6 (HID:0 DID:0) VID:19 (not sure if they exist here)
[cpu1] [SLFM]    FID:6 VID:19 
[cpu1] [IDA]     FID:12 VID:43 
[cpu1] [CURRENTLY ACTIVE FEATURES] IDA:64 EIST:1
So, looks like 19 is the lower limit for the VID. Pity.

That changes my table to be as the following:

Code: Select all

                B E F O R E  O P T I M I Z A T I O N           A F T E R  P H C
Freq MHz       VID    Voltage(mV)         Power(mW/ß)              VID    Voltage(mV)        Power(mW/ß)
1000            19         1004                 1008                19       1004                    1008
1333            27         1132                 1708                19       1004                    1344
1666            35         1260                 2645                20       1020                    1733
2000            43         1388                 3853                30       1180                    2785
Which is still good, but sadly I can not lower the Voltage below the standard :( for low freq. Is there a way to do so? How can be that during the VID tests the system actually crashed? That means that it really scaled down the voltage, or am I missing something?

Cheers

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

Re: Success T7200 2GHz D820 Laptop,Ubuntu 9.10 HALFPOWER

Post by the-fallen » Sun 8. Nov 2009, 18:42

We do not know if its possible to do that. I do not think so. They are calibrated during the facturing process (depending on some things like the quality of the waver) and those values seems to be solid in the CPU.

They also might differ on each CPU of the same type. I once had an T7200, too and as far as I remember I had a lower default VID (15 I think). But I am not sure about that.


Why it crashed? hm.
Did you have used an automated script? May it be that it crashed as it reached "-1" ?
The current available patch does not check if the value is < 0.

Post Reply