The "Undervolting" Guide

You don't need IDA..just uncheck it and don't worry
Anyway..this is strange, don't know why happens..

 

First off, Vista or XP ?
Secondly, uncheck that IDA thing, it's not really working for most computers, you'd have to be close to full idle to shutdown a core or run at different speeds.
That's up to technology to fix.
Make sure you ticked the "PERFORMANCE ON DEMAND" here:

 

I'm fine running without IDA, Its just weird why it does this, Just thought I would ask, Just in case it was something wrong.

I'm running Vista 64 bit, I have everything checked,, That I think need to be.. I have triple checked.

Guess, I will just leave it alone.

Thanks.

 

Can someone explain me how I can go lower of the default 0,900v on 6x multiplier?
My cpu is a t7500

 

under advance cpu settings-> intel core enhanced low power states

what does each option mean? which should we check off?

thanks

 

you won't be able to go lower than that

 

Yeah, Intel locked down the voltages on their Core 2 (and likely future) CPUs. Well, that's one reason to go AMD...

 

Those wont lower your voltages. The low power states are the sleep states you CPU goes in. No need to touch these settings unless your CPU makes a 'whine' noise.

You mean multipliers right?

 

No, I mean the lowest voltage that the processor can use. My T8300 is locked in at .95V while my TL-50 was entirely open (though it is a different generation and technology).

 

what should the settings be though?

 

With IDA, it will force the CPU to use a certain voltage. I did some tests before, for mine, with IDA ticked, it forced the multiplier under IDA to use either 0.1000V lower than default IDA or whatever voltage you have set if you set something lower than default for the IDA. I guess it's a stability precaution thing.

The only way I see how you can force the multiplier under IDA to use the desired voltage is, set the IDA the same as the multiplier under it, but I don't think that's a good idea.

Anyway, IDA comes on very rarely in real life situations, so I just left it disabled.

 

Yeah Intel does locked down the minimum voltage. However, from what I've seen, I found that the BIOS/motherboard play part of it too. For example, with the T7500, for some people, it is locked at 0.9000V, but some is locked at 0.8500V, they're not using the same model of laptop.

Either way, it is still not the reason to go AMD.

AMD uses a different technology, their CPUs will(should) never run cooler than Intel's. But I guess maybe I shouldn't say "never" with technology things.

 

Note that different samples of the same CPU model number are hard-coded to run at different voltages, according to the results of testing during production. It is not a feature of the BIOS or mainboard. It is a matter of luck what voltage range your CPU has been set to use. The Intel datasheets list even wider ranges than what we normally see in the products.

John

 

I just installed a pentium M 725 1.6ghz into my compaq.

I managed to drop the max multiplier voltage by .350v from the default 1.340v . I was able to drop the idle voltage from .988v to .700v

The latest CPU-z seems to be reading the voltages correctly for dothan processors.

Might do the OC pin mod if i have the time to pull it apart again..

 

Surprisingly...the school filter doesnt seem to block this thread...

Anyway, flipfire I have the same CPU in my compaq (Presario B3800), my voltages are...

6 - 0.700
8 - 0.764
10 - 0.892
12 - 1.036
14 - 1.116
16 - 1.164

I am sure i could push it lower as I've done it before, but after a few reformats i lost the settings... and havent really had the time to do it. I guess i'll see how low i could do it tonight... Also, lucky you! You have the right mobo to do a pin mod...i don't.

 

Next time... *cough*proxy*cough*

 

*cough* they are all blocked *cough*

PLUS I've got exams!!! What the heck am I doing here?!?!?!?

 

So, from what I've read, there's no way to make my Penryn go under .950 volts? Because almost all of my multipliers (x11 and down) are at the lowest voltage possible. :\

 

supposedly it's hard coded on your cpu. even if you some how set rmclock to a lower voltage it'll still run at .95.


what were u able to lower your 12x to?

i have my 12x set at 1.0v, i'll try to get it lower on the weekend when i have more time to run more tests.

 

Ah, that sucks. :\ Oh well, at least it's pretty cool now.

I'm currently doing 2 hours long test with x12 .9625. I tried .9500 yesterday but got error message 6 secs in.

 

X12 at .96V? wow, cannot go lower than 1.05 in my case...

 

What are you using to test, Orthos? I tried OCCT but it didn't give errors and occasionally I'd get weird BSODs (rarely though), however, with Orthos I had to bump up the voltage another tick to avoid errors and haven't BSODed since.

 

damn that's really low

i'll try testing on 12x 0.9875v on the weekend

 

rtrdogs managed to get 0.925v for all his multipliers on a T8100 (vostro)

And if i remember correctly the lowest voltage for his T9300 aswell. It really impressed me how low penryns go for the highest multiplier..

 

Yeah, .9625 is working well for x12 after 2 hours test. So, it's set there. Rest of the x's are .9500 (the lowest it can go).

Aw man. Does that means the limitation is from the MoBo, not the cpu itself? Oh well. It's pretty low as it is, so I should be happy.

 

what kinda temperature readings are you getting?

 

It's hovering around 57- 60 degree Celsius and less after the two hour test. In normal use it's around 40 degree Celsius.

 

what was it originally before under volting?
thanks

 

It was about 71-73 degrees C at 1.1375 (default).

 

Have recently undervolted a M1530 with T7250 and managed to achieve a temp drop of around 10 degrees under test conditions. The volts I have managed are as follows;

SuperLFM 6.0x - Stock 0.9000V
6.0x - Stock 0.9000V
7.0x - Stock 0.9750V - Stable 0.9000V
8.0x - Stock 1.0375V - Stable 0.9000V
9.0x - Stock 1.1000V - Stable 0.9375V
10.0x - Stock 1.175V - Stable 1.0250V
IDA 11.0x - Stock 1.2750V (Disabled)

Everything appears good withrespect to stable operation and cooler temps etc.

However, two concerns i have which i don't have an answer for after reading the whole of this thread. Firstly, when i run PCMARK05 benchmarking a get a reduction in result with rmclock enabled as current and set POD against rmclock open but current set to N/M about 5000 to 4900. Therefore a performance drop.

Secondly as i type this on a separate desktop the laptop is sat at idle with only rmclock running and clock speed @ 600 Mhz, the CPU load is consistently between 50-60% whilst OS load is around 2-5%. Why ???

Any ideas guys

edit: Also forgot to add in a very simplistic battery test of running ORTHOS on battery at high performance settings the POD settings achieved a 1h 12m run to 15% battery life whilst with no management a 1h 1m run to 15% was achieved so an improvement was seen. Obviously need a proper power benchmarker to get accurate results !!

 

Yeah, it's a real shame that it wouldn't let the voltage to be changed below 0.95V. At this rate it seems like it could easily do around 0.8V for the lowest multiplier.

BTW, for those of you with T9300, do you use SuperLFM and IDA? IDA seems to be working fine on mine, but when I turn SuperLFM on, it creates weird CPU spikes (especially noticeable when listening to music, which starts skipping every few seconds). Turned it off for now, but would be nice to go into a lower state.

Another thing I was wondering about - when undervolted, the core temperature fluctuates from ~35C to 45C when the fans start spinning. Would this cause considerably more wear on the CPU from expansion/contraction than if it was in the non-undervolted state where the temperature varied only a few degrees (42C-45C)?

 

Hey man, same laptop and same CPU here. And same results. I also get a SLIGHT score drop in PCMark, wPrime, when running RMClock. I also get the 50% CPU load but small OS load when I am on LFM. Do me a favor, go into Device Manager and just disable your soundcard (if you don't know how to do this, let me know and I will walk you through it). I bet you your CPU load will drop. I have been trying to find a solution for this, but have not been able to.

Also what temps are you getting max load? I just did a 6 hour Orthos test yesterday, and my max temp was 62c.

 

Woohoo, you guys make me want a Penryn BAAAAD.

PCMarks score might drop due to RMClock using a bit of the cpu plus some ram, although it can't affect too much of it.

 

Re: Benchmarks, RMClock and P-state transitions.

I've found very significant benchmarking differences when enabling pst. I get comparable performance when my CPU runs flat out, however, with any kind of "spiky" benchmarking (see: 3DMark) it gets ugly.

In 3DMark, for example, I lose around 2000 "marks" with it enabled. My suspicion here is in the step up/down time. Mine are set low - 200us - but it's still noticeable.

I'm going to be experimenting with a quick step up time and a significantly longer step down time (likely in the 1-2 second range) and see how that effects things.

Basically, I want to get the extra battery life out of idling at low speeds, but I want to get 100% performance out of my CPU for anything else.

 

I don't know. But I noticed that you are OC'd. I would still think that lowering the voltage on a OC'd CPU, could be a possible cause of problems.

 

Nah, I ran stability tests while the CPU was OC'd when determining minimum voltages, and I actually run a bit over the lowest stable voltages to give a bit of a margin.

That said, the performance loss is noticeable while OC'd or not. Seems to me that as the CPU isn't under stress continuously but rather in spikes (this can be seen by watching the load monitor while 3dmark is running) it's spending a lot of time overall actually running "downclocked" before stepping up the speed. As every break in load results in a drop in speed, followed by (in this case) a 200us delay before ramping back up again, that those "nibbles" add up.

Until I get it figured out and confirmed, though, I'm just running with my CPU at full speed while under AC power and using PST while on battery power. The battery drain difference is really very signficant.

 

You shouldnt be losing that much points. Something is definately wrong.

In management try changing P-state transition to Perform Single step transitions only

 

Ooooh, here's something, too.

When I'm on battery power, I seem to lose all my RMClock settings, and they revert to defaults. On closer inspection, it turns out IDA is completely disabled and my 13x multiplied (VID 8) simply doesn't exist. This caps my performance under RMClock somewhat immediately, as the processor is throttled to 12x max, instead of 12.5. If I plug the notebook in and restart RMClock, IDA can be set again and the 13x multiplier is again visible, but all settings are lost.

That's really annoying.

 

That explains the pts drop on the benchmarks. Your handicapped by 100mhz

So your IDA becomes disabled only on battery power?

 

happens with me as well
i'm using power4gear, if i select "maximum battery saving" under "process control" it'll disable IDA

 

This actually make sense. I would expect from a battery saving software to disable(at least) the highest multiplier. Don't know if they're supposed to do that but makes sense

And if I have 2 or more battery saving programs I would expect some issues when running them simultaneously.

 

Actually its only disabling IDA for some reason...and T9300's need the 13x IDA multiplier to use the 12.5x multiplier .o well you shouldnt be using your 12.5 multiplier while on battery anyway... no need for that much power while on battery.

 

ya I'm glad I can disable IDA through power4gear so it can save energy and lower the temps while I'm on battery mode.

i'm currently running, power4gear, vista battery saver, and rmclock together. things seem to be stable so far.

 

Weird, I just tried it and even on battery , IDA stays there .

But then again, I use a 6x SLFM so...

 

I use a 6xSLFM as well. Strangeness. Thing is, I *want* my lappy to be able to ramp up to 100% (actually, beyond, thanks the the FSB OC) because I don't do a lot on battery power, but I like to be able to show off if you will when I do it.

What's REALLY strange is that RMClock, on my woman's Dell, shows her processor running at .5x multiples. Of course, it fails to detect her horrible little ancient AMD processor at all, but still... I wonder why it shows .5 multiples on hers but not on mine?

 

Am I correct in assuming that undervolting will do little to change the cpu's temperature when idle? Flipfire's instructions say that you should not adjust the voltage settings for the base multiplier (which is the cpu's state when idling), so adjusting the other multipliers should not affect the idle temp, correct?

 

My CPU used the highest multiplier when idling, so it did lower the idle temp, since it's now using the lowest multiplier.

 

I guess it depends on your particular cpu and how much it's doing in the background. For mine, the cpu appears to already be on the lowest multiplier when idling, so I would expect undervolting to have little effect for the idle temp. Undervolting would still be effective for loaded states, of course.

 

For idle, UV won't change much in temp, mine dropped 2-3c when using SLFM for idle. The main temp benefit of UV is the max cpu load temp. For the base multiplier, I didn't have to adjust anything, cause by default it is already using the lowest it can be selected. But not every CPU is the same, you should still test it and find the lowest, some people are able to go to 0.8500V for their lowest multiplier, which will probably drop idle temp a bit more than 2-3c.

 

Asuming you only tamper with the max multiplier it won't.
If you lower ALL voltages it will.
My stinky 2.5 year old HP' T2300 idled at 55-60C before,and after some lowering, idles and 45-50C and maxes out at 70-75C instead of 85C.