The "Undervolting" Guide

[DanSpree]

Very nice!! Thank you!

I set it to 48, and it's showing me 72x - since 48 is hex for 72, I know this seems correct.

Doing a Prime single-core test, I'm running around 8.40x - 8.45x or so, and in this mode I'm getting 2240 MhZ.

This is where RMClock breaks down. My cPU is at 111% load! But I know all is good from the Throttlestop window.

In RMClock, I set the voltage for IDA at 1.075, just to be safe... Since this is pretty much untested. I may run the single-core test at 1.05 for an hour or so, just to see how it does.

I really can't thank you enough!

[unclewebb]

I just uploaded ThrottleStop 2.75 so hopefully that correctly recognizes your Q9000 as a mobile CPU so the VID voltage will work correctly. Is the minimum the ThrottleStop VID adjuster can go only 1.05 volts?

When using half multipliers, RMClock gets a little confused. You can trust the multiplier that ThrottleStop is showing you.

You will never get the full 8.5 multiplier because there are always some background tasks that need to be processed. As soon as a task wakes up the second core, the maximum multiplier drops down to 7.5 for both cores on the same die. When the task is finished, that core goes back to sleep and the multiplier goes back up to 8.5. This happens hundreds of times a second. There are two dual core dies within your quad. That's why the average multiplier is reported slightly less than the full 8.5.

Another interesting test would be to run two threads of Prime95. You could then do the same with SetAffinity... and lock them to different sets of cores. That might be interesting. If you did it right, you should be able to see Dual Dynamic Acceleration in action with two cores close to 8.5. If you put both tasks on the same die, the multiplier should be at only 7.5. Hopefully the OS and CPU are smart enough to schedule tasks appropriately to maximize performance.

Thanks for your feedback. I finally had the chance to learn about the Q9000. The Intel docs and the internet are short on information about this CPU.

[jupapri]

yes. for quad cores the minimun voltage is 1.05 volts, i don't know if there is a way to undervolt to 0.95v

[DanSpree]

It is showing a different voltage, and I think a couple checkboxes that were selectable before are now not... so it appears to have recognized something - Here are the screenshots (1st on Idle, 2nd at single-core test on Core-0.)

[DanSpree]

yes. for quad cores the minimun voltage is 1.05 volts, i don't know if there is a way to undervolt to 0.95v

Ah, thanks. Well, I'm satisfied. All this work has paid off completely.

[unclewebb]

The voltage formula is different for Core 2 mobile CPUs and the SLFM box is now unlocked in version 2.75 like I think it should be. When your CPU is idle, does having SLFM checked drop the bus speed or total MHz?

Is the lowest voltage 1.05 volts in ThrottleStop 2.75 now? That info is being read from the CPU so if that's as low as it goes then that's as low as it goes. There won't be any software that can make it go lower than that minimum.

Almost forgot. The new Temp button should convert the DTS data column to the core temperature.

[DanSpree]

Hmmm, no... Is Voltage ID what I'm looking at? It's 1.2625 regardless of if SLFM is checked or not (even after hitting "Save"). Could this be a problem with RMClock? Maybe your voltage is correct, and RMClock failed somehow?

On a side-note, now that I know about the task-manager way to have a program use a certain core, would it actually improve performance to force a game onto one of the back-end cores? Or movies that are sometimes choppy, if I can put it on Core 2 or Core 3, I wonder if it would force an IDA state.

I bet the CPU does that anyway, but next time I have a choppy movie I'm going to try that out.

[DanSpree]

Wait a sec, I see what you mean - Bus speed... Using OCCT, on idle, my bus speed is 266. It stays at that value during Idle when I check SLFM

[unclewebb]

With the adjuster in ThrottleStop, you should be able to adjust the VID voltage up and down. Just click on those little up and down arrows to the right of the 1.2625 number to play with it and to see how low it goes.

Do some Prime95 testing where you start two threads of Prime95 and then lock it to any two cores. You should see what two cores is going to end up with the highest multiplier. You could try running a game locked to those two cores then but it might not get you any significant difference.

To test SLFM you would need to turn off RM Clock and then you would need to turn on ThrottleStop with the Turn On button so it shows that it is enabled. You would also have to check the Set Multiplier box and the Power Saver box. With your CPU idle, that will show if the bus speed drops down or not. I'm going to check the Q9100 docs to see if this is supported.

Edit: Now it makes sense. The Q9100 does not support SLFM mode so that also explains why the lowest voltage is probably 1.05 volts. The lower voltages like 0.85 volts and 0.95 volts were only used on CPUs that support SLFM.

[DanSpree]

EDIT -

Okay - so I performed those tasks - I turned off RMClock, Turned on ThrottleStop, Checked the Set Multiplier and brought everything down to min (6.0x @ 1.05 V). Bus speed remained a constant 266 MhZ throughout, regardless of the SLFM checkbox, so it may not be supported, or I might be doing something incorrectly.