Undervolting a T9300!

Hey,

I am trying to run a T9300 at its stock speeds with as low volts as possible.

Currently I have dropped the max VID to 0.975V for 2.5Ghz and its running stable. Mind you, the default vid was 1.13v (its a 45nm 6MB L2 Cache CPU) and because of the decrease in volts, battery life has gone up and the heat being generated at full load has gone down significantly.

Should I keep trying for more? I wish I had a frame of reference to work with.

The minimum voltage allowed is 0.925V which was by default being used for the 100x8 mode (where the CPU halves the FSB and bumps the multiplier from 6 (minimum with EIST) to 8 to maintain a 800Mhz minimum clock speed.

Its perfectly stable at 0,975V at 2.5G (200 x 12.5) and the reduced heat output is a boon.

Karan

 

Basically, its all done from within Windows with software (RMClock to be specific.) It allows me to adjust multiplier for various multiplier and processor states (like DFS and IDA) I am guessing that if it is not stable then the Orthos session will fail or at max I will get a BSOD (no kind of corruption as the RAM is totally stable and not being oced...)

I have RMClock doing all that for me I don't have to manage voltages myself... I just hv to test and set (and Save the settings)

I will post screenshots of the settings when I get home. RMClock is a gem of a software.

The 0.825V ULV processors run at max around 1.3Ghz. I am trying to stabilize almost double that clock speed. I will definitely try and see how close to the minimum voltage (0.925V) I can get.

 

how did you get 12.5x as the current RMClock only allows 12x.

 

Have you run an extended stress test? I also have the 9300 and have similar results (1.0V at 2.5G) so I wouldnt be concerned. I got BSOD at .975v so I bumped it up to 1.0 to be safe.

 

I make my test this weekend, and got this results

At 12X

Test V Temp Time of test
Original...............1.1375.................80.........45 min
Test 1 ...............1.0875.................77.........45 min
Test 2................1.0375.................74.5......45 min
Test 3................0.9875.................65.5.......45 min
Test 4................0.9625.................BSOD...... 1 min
Test 5................0.9750.................BSOD.......5 Min
Test 6................0.9875.................67...........3 Hours 1 min

i Got a total undervolting of .150 V and reduction of 13 C wich i think is good.
still i have to change ther rest of the multipliers

i an using the 1.000V for now not risking geting a BSOD

 

1v for 12x is what most people get. I want my 12.5x multiplier!

 

These are my current settings. Testing for stability. The drop in CPU temps and the overall heat is astounding.

Mind you, RMClock doesn't seem to support half multipliers, hence, I have kept IDA enabled which actually takes care of the 12.5 mutliplier (200 x 12.5 = stock clocks.)

^^ Thats how.

Basically RMClock uses a > rule when applying p-states. So if you have IDA enabled, it allows you to set the VID for the 13x multiplier and that is the state which is used for the 12.5x multiplier on the T9300. IDA is engaged when there is a single core load, and thus it doesn't require that much voltage to remain stable.

See my screenshot on a reference on how I did it.

 

Which test was this? Good drop in temps bro.

 

IDA only runs 1 processor though, so you never get dual core at 2500MHz:

http://www.lenovo.hu/kszf/adatlap/Prosi_Proc_Core2_Mobile.pdf

 

the 12x runs at 2.49GHz anyway

 

I run at 12.5x when I plug in my AC power. I find it is still clocked at 2496Mhz according to my RM Clock. It won't go anymore higher than that.

 

And what does your throttle read as, or do you just bypass the undervolting when on AC?

 

12x200 = 2400, so no, it doesn't.

 

Wait I get a screenshot.

EDIT:


Damn I got 1.1375 volts and it's reading at 12x... So much for my fat hopes... Yeah that is, on AC. I think most probably it doesn't get up to 12.5... I think. I rarely run RM clock only just to remove the CPU whining.... Any help on undervolting wihtout the loss of performance is appreciated thanks.

Anyway I did set a profile "performance on demand" then selected the multiplier to 12.5x on P-State Transitions.

 

You're not running RMClocks voltage management which is why you get the 12.5x multiplier. It makes sense now, thanks.

 

I am buddy... trust me on the IDA thing. Just set whatever voltage you want for dual core load to the 13X multiplier and it will be used with 12.5 dual core load.

I will post screenshots later when I get home.

 

I set IDA like you said but even under orthos full load it never goes to the IDA setting.