The "Undervolting" Guide

Another thing came in my mind. I think maybe, I don't know, maybe it's the higher expansion range in temperature Intel doesn't want. Although using SLFM, I'm seeing about only 2-3c difference at idle.

By the way, I'm using SLFM also and it's set @ 600 MHz.

It's not normal, typing shouldn't cause CPU to load up to 56%. Open Task Manager and see which app/process is actually causing the CPU load.

 

hmmm, how much of a drop can you expect when you undervolt a penryn? I ran orthos for 15 minutes and got to 68.

 

Temperature wise, id guess around 6c cooler. Its not much, but since penryns seem to be able to reach low voltages, you will see more battery life.

Give it a try and see for yourself..

 

I was able to drop mine down to 58C max

 

That is the thing, Task Manager does not anything, at all. Hence CPU load shows 50% but OS load will show like 2-5%.

Just at desktop, does the green bar (the left most of the two) show any activity?

Oh this is at the lowest multiplier (6x). Of course the higher the multiplier goes the cpu load goes down.

 

Strange, normally in Task Manager, you can see under the "Process" tab what each process is using in CPU load.

At desktop, both of my green bars are at 0, the left one sometimes jumps to 1, but rarely. Even now that I'm typing in in FF, same result. I'm using SLFM and has it set at 6x lowest also.

 

Hmmm this is troubling. I did some research on the RMclock forums and some other users have this issue too. But it seems to be with old chipset. What I read is if you disable the USB powersave function it fixes it, but from what I can tell (unless I am looking in the wrong spot of course), this function is disabled.

My PCMark is good (5.1k). So I am not sure if it just a reporting bug, or if my cpu load really is always high.

 

I believe you have the same chipset as I do.

Have you tried using the latest chipset driver directly from Intel? I did not use the version provided by Lenovo because it is not the newest. I used the driver directly from Intel.

http://www.intel.com/support/chipsets/sb/CS-022768.htm

 

Do a benchmark using PCmark and wPrime.

First do a benchmark with RMclock turned OFF.

Second test will be with RMclock ON.

If your marks have dropped then the CPU is really being used by another program. Do the test 2-3times just to make sure. I suggest you void the first test you do as its a little innacurate (cpu needs to be like warmed up)

 

Thanks for the great guide. Can I still use Vista's default power plans (high performance/balanced/power saver) or do I need to modify/turn them off somehow for this to work properly?

 

I did quick wPrime benchmarks.

wPrime is a benchmarking application designed to use a highly multithreaded approach to calculating the square-roots of large amounts of numbers (up to 32 billion at this stage!)

Results as follows:

WITHOUT rmclock: http://img293.imageshack.us/img293/9122/withoutrmclockcv1.jpg

I got a base score of 37.343 seconds. The lower the time the better. This score is right on target for a T7500 CPU. The max temperature reached when doing the benchmark is 56c

WITH RMclock: http://img152.imageshack.us/img152/4284/withrmclockde9.jpg

Suprise Suprise. I still got the same score of 37.343 seconds but this time notice the max temperature reached is only 50c when doing the benchmark. This is a clear sign that undervolting is active.

This test proves that undervolting doesnt affect performance. The only way it will change performance is if RMclock isnt setup properly.

I closed uneeded programs as possible and i reset HWmonitor and wPrime between the tests to increase the accuracy of the results. Also i voided the first set of tests just to get the CPU into operating temperatures.

I might do the PCMark benchmark later on but i cbb'd downloading it atm. Ive done the 3dMark06 benchmark before and it was also the same results on the CPU score.

 

Did the "Undervolt". Thanks, works great!

The only thing that bothered me a little is that I spent about 4 hours to find out that my lowest possible voltage is the lowest selectable 0.925V from 1.1375V. Temps are about 14C cooler. I'm very happy with this though I don't really know how much battery time I've gained....

 

flipfire, actually by default, from personal testing, RMClock doesn't override Vista's power plan completely unless you tick the checkbox in (RMClock -> Profiles -> Sub Profiles -> OS settings -> Use OS power scheme -> Custom). I just left it unchecked, which is by default anyway.

For me I didn't have to choose High Performance before setting up RMClock. Maybe different manufacturers come with different PM software. With Lenovo, I just kept at Balanced and everything is ok.

This is what you're talking about, right? Or it's something else?

 

Yes it doesnt take over completely, sorry what i meant was CPU management not power managment but yeah you can set RMclock to take over vistas power management too

Im just talking about the profiles, even if you set it to 'Power Saver' profile (which is supposed to cut your CPU speed in half) RMclock will just override it and use its own cpu management.

 

Flipfire, thanks for the benchmarks with and without undervolting,,,

hopefully they will put a few questions to rest,,,

bigO

 

my max temp is only 63 C it went up to 60-63 then the fan kicked in and it stayed stable at +/- 55 is it even worth it to undervolt ?

 

No problem bigO

crazyanz: Well if your really satisfied with your system at the moment, then leave it be. Doing the stability tests can take hours so why bother if your happy with what you already got.

Undervolting probably wont change your temps much since your CPU is already cool to begin with but it will definately increase your battery life

 

the RM cpu tool doesnt even recognize my cpu it does see it and measures it but at cpu info it says unkown core. it does see its a C2D

 

Thanks for the info. In terms of results, I have a Macbook Pro with a T7500 (2.2 ghz) processor. Before undervolting my max temp was 92C After undervolting (1.2375 --> 1.0375 at highest multiplier) my max temp was 84C, which is a 8C improvement. Interestingly, though, my "average" temperature (where it hovers at during testing) seems to have barely changed, from 80C originally to about 78-79C after undervolting.

I suspect it may be the GPU that's contributing to the heat, as it's consistently hotter than the CPU temps. It's an Nvidia 8600 GT. I may have to look into undervolting or even underclocking it. Anyone have any suggestions? Overall, though, I'm very happy with the results. Thanks again!

 

I knew MBP's were ovens but I didn't know they were THAT hot.

 

If you think a MBP is an oven, then a MBA would be a BBQ...

 

Temperatures also depend on the capacity of the cooling system and the temperature settings for the fan operation (which are defined in the BIOS). Some notebooks are designed to run hotter than others with less fan operation. My Zepto has integrated graphics and a cooling system designed for dedicated graphics so, with undervolting, the fan never gets above its lowest speed and the temperature never exceeds 70C. Without undervolting the fan would, under heavy load, go up to the next speed setting.

John

 

Hi!

After the recent good weather my room is about 28C - which is pretty annoying for my fan and my CPU (Pentium-M 760 2GHz, not overclocked) is getting way too hot (up to 75C within a few hours under full load).

Now I started undervolting with Notebook Hardware Control (basicly works like RMclock).

Stock voltage for 2GHz was 1.356V - I got down to 1.086V and tested with Prime95 (Small FFTs and Large-InPlace FFTs) and HOT CPU Tester for about 90-120 minutes. Then I started playing a game which also uses 100% of my CPU capacity and nearly no GPU (Hearts of Iron 2 btw.). CPU temperature was 60C after 2 hours and GPU 65C.

Suddenly my notebook screen goes black and I have to do a hard power reset - so even if Prime did not give me a single error, no blue screens is that a result of undervolting or could it be something (coincidentally) else?

Coz I thought normally those calculation programm give you errors if you are too low and only if you are way too far beyond your voltage it just goes out.

Hope I did not forget to mention anything else important. Btw. Hi and excuse me for my rather bad english (Austrian guy here

 

Actually you should start just below max voltage and test lower and lower voltages until you have a failure....

now that you have done that,,,, if you have lockups or strange stuff happen, then it sounds like you need to raise that voltage a little and try again....

read this guide for great info and details on how to stress test,,,, and if you experience lockups, BSODs, or any strange activity then you should raise voltages more...

bigO

 

Well that is the funny thing - stress testing for nearly 2 hours did not give me any lockup, blue screen or strange behaviour at all. It suddenly appeared while playing a rather old 2D game and it only happend twice so far but never during a stress test, those always worked without any errors.

From 1.086V to 1.064V I got a reboot within a few minutes - at 1.086 it seems rather stable at least while running Prime95 and Orthos at the same time. So I am a bit confused because I always thought before I get a BSOD or something similar Prime95 would come up with a bunch of errors due to misscalcuations.

 

Use Orthos, it`s a far better CPU stress tester nowadays...

 

Ok so I posted a few posts ago, that I had a contast "CPU Load" of about 50% but OS load of 0-5% @ idle.

So I reinstalled XP, and installed each driver one by one. And guess what? It is my damn Audio Driver for my sigmatel card. Now I get to figure out what the heck to do about it. yay

 

powz: You might wanna invest on a notebook cooler for your MBP

cmdrcool: Set it to 1.100v just to be safe. Try using RMclock instead of NHC to undervolt. The screen going black never happened before, i can only assume its something NHC did.

Tusin: Try using a different version. It might have something to do with SP1 since sigmatel was listed as a problematic driver

 

Actually I reinstalled XP Pro. I am trying to find more versions, but so far I am having the same issues.

I don't expect people to know about this "issue", because most would look at OS load and not CPU load. As in Vista/XP reports the OS load as the CPU load.

 

Have you tried updating the BIOS from Dell? I'm suspecting it's caused by the BIOS, because I'm running dual-boot Vista SP1/XP SP3 and no problem in either.

 

Yeah I have A08

 

Ok I did some WPrime tests aswell. And I did follow the guide to the T. But if someone thinks I made a error, let me know what to look for

T7250

Did 4 runs of each:

RMClock On: 43.859
RMClock Off: 43.249

So with RMClock off, it is "faster", but not enough for me to care about it.

I also tested with the Sigmatel Card on and off.

With it on I am getting a 44.249

 

Void the first set of tests. The first test i did without rmclock, i got like 38.5xx when i did it again i got 37.xxx which was a little inconsistent

Its because the CPU needs to throttle from SuperLFM 6x to 12x which causes the slight latency

 

I didn't count the first set of tests. I also ran it with it only locked @ 10x (my hightest multiplier), and there was pretty much no change. That does tell you that RMClock does a really good job changing multipliers.

 

Weird, your wPrime score should be spot on or at least mariginally .250ms away from the base score. Make sure you close as much uneccesary programs as possible.

I did the same tests on my other dv2000 (T2500 Yonah) and it was under .250 away from the base score.

Im not sure why my T7500 is spot-on everytime. To be honest i was expecting it to be at least .050ms slower with RMclock on. I think i may have to do the PCmark tests as a better benchmark

EDIT: Okay i re-did the test with RMclock and got slightly higher which is very marginal. http://img405.imageshack.us/img405/3981/withrmclockob3.jpg

 

Anyone uses RMClock Pro here?

 

Ok well I will double check my settings. I am going to reinstall XP again tonight (with all the system restores I was doing trying to figure out why the Sigmatel card was increasing load @ idle so much, I am sure I fubard something).

 

I was able to lower the 12x multiplier by 0.125. Is the best way to do the rest of the multipliers the same way? If so I'll be at this for a while XD

 

I think that its pretty safe to assume that a lower multiplier will use at most the same voltage as the multipliers above it. At least that's how mine were.

Someone should confirm this though.

 

W00t! Got my reUnderclock done ran Orthos on priority 9 for 8hrs no problem and drop my temps from 69C to 56C!

 

Not bad. I dropped mine from 79 to 61C

 

Could an unstable undervolt cause the screen to start flickering after playing games for a while, or is that more likely a driver issue?

 

If its flickering/flashing like 10 times per second then that would be nvidia graphics card. Specifically the 174.xx drivers are known to cause this.

To make sure, turn off RMclock for a while and see if it re-occurs

 

Wow, read that post then experienced it 10 minutes later, damit. Asus told me to update video drivers as it was the supposed solution to my eternal bsod issue, now more issues. I really liked the 174 series, now I gotta go find some other series to use, gah, more graphics driver testing, possibly the thing I hate most in life.

 

My screen used to randomly freak out with the 174.32 drivers. nvdllvmk.sys causes crashes. Im currently reverted back to the rock stable and OC'able 169.04

I hear the 175.xx is quite a treat.

 

the weird thing for me is that I can play other games fine its just guild war that it bugs out with, oh well.

But I was able to lower my temps by about 5 degrees, now I need to undervolt all the other values XD

 

I am going to try 169.28, 167.49, your 169.04, and then maybe the 175's, whew, quite a Thursday coming up. TY for your input.

 

So, will these settings be absolutly unique for every machine, or is it possible to have people post their results from their CPUs, and then we can use those settings if we have the same CPU in our machine?

In other words, if I have a T8100 Core 2 Duo, could I just use the same settings as someone who went through the testing procedures on their T8100, or do I need to run all those tests myself?

In other-other words - can I just be a big, lazy mooch, or do I need to work at this thing?

 

It's unique to each machine, some were able to undervolt more than others. Another thing I noticed is that even the same CPU, different bios/motherboard can have different results.

 

No, im afraid your gonna have to to take the time of doing the stability tests. CPU's arent built equally. Each chip has its own voltage tolerances. Some can go lower than others.

The stable voltages other T8100 users got are possible configurations, it doesnt exactly mean it will be stable on yours. It should only be used as reference.