The "Undervolting" Guide

I can't think of any reason why it should be. I deliberately programmed it so even when logging, the log file is only written to the hard drive once per minute. When not logging, ThrottleStop shouldn't be writing anything to the hard drive.

I'll definitely look into this because I hate software that accesses the hard drive when it doesn't need to. Are you using TS to monitor GPU temps?

Edit: I have an ATI GPU. When ThrottleStop is monitoring GPU temps, there is a pile of hard drive activity. As soon as I disable that, there is zero hard drive activity. I guess you'll have to complain to ATI since I'm asking their driver for the GPU temperature. Typical ATI code bloat.

If ThrottleStop has the Log File option checked then you will see the once per minute log file writing as I mentioned before. There will also be some hard drive activity when the program first starts as it reads the settings as well as when TS exits but that's all the hard drive activity that I could find. I was pleasantly surprised actually, except for the crappy GPU code but that didn't really surprise me. Other than for testing purposes, probably better turning that crap off. Let me know what you figure out. I'm more anal than most when it comes to disk thrashing and avoid it as much as possible.

Edit2: ThrottleStop by itself doesn't touch the registry. I hate that thing. Asking the ATI driver what temperature the GPU is at results in 400+ accesses to the registry every time the temperature is updated about once every 5 seconds. Pretty pathetic.

 

I don't have GPU temperature monitoring on neither logging... And I also have a nvidia gpu.

I use throttlestop with "set multiplier" "power saver" "slfm" "bd prochot" on, minimized to taskbar.

 

My laptop has an Nvidia GPU but for this test I have GPU temperature monitoring turned off. This is with ThrottleStop open.



After 10 minutes and about 100,000 events, not a single event was caused by ThrottleStop. Nvidia GPU temperature monitoring will cause regular access to the registry. Once again, I'm just asking the Nvidia code for the GPU temperature. When GPU monitoring is disabled in ThrottleStop, there shouldn't be anything showing up in Process Monitor other than when ThrottleStop starts up, ThrottleStop exits or if you open up ThrottleStop after it has been minimized to the system tray. After it has opened up, there is no access to the hard drive going on that I can see.

Can you tell me how you have Process Monitor setup. Am I missing something? I've tried to duplicate your settings but I'm not seeing any disk activity. I've tried this on my desktop which has an ATI card and on my laptop with an Nvidia GPU.

Here's the version I'm using.

ThrottleStop 2.00 Build 26
http://www.fileden.com/files/2008/3/3/1794507/ThrottleStop.zip

 

unclewebb: looks lite a neat little program you've made but how exactly do I use it, any basic guidelines, I suppose I have to check "clock modulation, chipset clock mod and multiplier and then set my voltage ID, but what settings do I use?

I'm a bit of a noob once again since the last time i undervolted was quite some time ago

 

Remove all the filters. Take a look, it's explorer.exe that is generating the hdd usage, querying stuff from throttlestop.exe every 9 seconds exactly. It only does that with throttlestop... That page has 20secs of monitoring my hdd.

 

TigTex: Thanks for showing me that. I have no idea why explorer.exe is constantly querying ThrottleStop. As I showed in my screen shots, ThrottleStop by itself is not initiating any access to the hard drive during normal use. It might be very difficult to track down what explorer.exe is up to but I will certainly look into this some more. I'm curious too.

e2zippo: You have to give me a few more details so I can help you out. What is your CPU model number and do you have a laptop? Most computers that use modulation will either use clock modulation or chipset clock modulation but not both so there is no need to check off both. You only need to check the one that is dropping down on your computer. Before checking anything I'd run a log file and test for throttling. No use fixing something if you don't have a problem.

To under volt using ThrottleStop all you have to do is put a check mark in the Set Multiplier box, maximize the multiplier and decrease the voltage until you lose stability. I usually do some brief tests with Prime95 Small FFTs when testing for basic stability and then when I see what the CPU is more or less capable of, I do longer tests after that. Some of the other settings depend on what CPU you have.

 

ThrottleStop 2.00 Build 27
http://www.fileden.com/files/2008/3/3/1794507/ThrottleStop.zip

I found a bug that I think was causing explorer.exe to check up on ThrottleStop on a regular basis. I'm not sure why but I found a way to work around this issue so it should be OK now. If you have any problems TigTex, let me know. Thanks for finding this.

 

Everything looks fine now. Thanks for fixing the possible bug.

 

I've got a Core 2 Duo P8700 and was wondering what a good starting voltage would be using ThrottleStop?

Looking at ThrottleStop's log when the machine isn't doing anything in particular the multiplier seems to rapidly switch between different values. I'm seeing numbers like 9.86, 9.75, 9.60, 9.61, 9.58, 9.54. It goes to 9.50 (and seems to stay put there) when I run a Prime95 test.

Why is this? Why is 9.50 seemingly magical when it's under 100% load? Does this mean that I should force the multiplier to 9.50 or 10.0 (which ThrottleStop defaults to) when attempting to undervolt the CPU?

Which of the other checkboxes should I check?

 

The P8700 has a default multiplier of 9.50. When both cores are in the active state, that's the maximum multiplier that you will see. This CPU supports Intel Dynamic Acceleration. When one core goes to sleep and enters the C3/C6 sleep state, the other active core can use the 10.00 multiplier. In Windows there will always be background processes waking up the second core. As soon as this happens, the maximum multiplier will drop back to 9.50. ThrottleStop is able to accurately calculate the average multiplier based on data coming from high performance timers within the CPU. When IDA is being used, it is normal for the average multiplier to be constantly fluctuating somewhere between 9.50 and 10.00. If you run a single thread of Prime95 you should see the average multiplier get close to 10.00 if you don't have too much other background activity going on.

If your laptop doesn't use clock modulation or chipset clock modulation then there's no need to check these boxes off. If you run a log file while fully loaded then you will see if either of these apply to you.

Every CPU is unique. Lower the voltage and only use as much as you need to remain stable. Run something simple like Prime95 Small FFTs and start lowering your voltage until you lose stability. When you lose stability or your computer locks up, you've gone too far. I don't waste too much time finding an initial setting and them after that you can run more benchmarks for a longer period of time to see how stable your reduced voltage really is.

A P8700 might be OK at 1.05 volts to 1.10 volts.

 

I tried forcing the multiplier at 9.5 (and nothing else is forced) and undervolting at 1.0750V at the moment. This is the log following a Prime95 Small FFTs and Furmark (stability) run:

05/29/10 23:24:53 9.50 64.3 100.0 87.5 26
05/29/10 23:24:54 9.50 73.3 100.0 62.5 24
05/29/10 23:24:55 9.50 61.5 100.0 75.0 28
05/29/10 23:24:56 9.50 69.6 100.0 62.5 26
05/29/10 23:24:57 9.50 61.8 100.0 62.5 28
05/29/10 23:24:58 9.50 61.1 100.0 62.5 28
05/29/10 23:24:59 9.50 48.4 100.0 62.5 29
05/29/10 23:25:00 9.50 58.6 100.0 75.0 28

I'm still not sure quite what these numbers tell me.

Is forcing CHIPM to be 100% the correct way forward (since that seems to be the problem area)? If I set the CHIPM value what should I be on the lookout for besides ominous smoke plumes from the 90W adapter I've got running the laptop? Does it matter that C0% isn't constantly at 100%?

Is the success of undervolting black and white or is there something in between "100% stable" and "it crashes every time"? Could the result be an ever so slight amount of data corruption for instance?

 

Looks like your chipset is throttling a lot... did you take a look at your system temperatures? forcing it to be at 100% all the time (when overheating) might destroy your system.
Maybe it's time to clean the dust from your pc. It's too hot!

 

The chipset clock modulation column says that your CPU is being slowed down internally (throttled) by the bios.

The number at the far right in the DTS column tells you how far you are away from the Intel designed thermal throttling point. That says you are about 28C away from being "too hot" so your CPU is nowhere near catching on fire. These CPUs do a great job of looking after themselves and even when running ThrottleStop, can still over ride any settings in ThrottleStop and slow down if they need to.

When the CPU is being slowed down internally by the use of clock modulation or chipset clock modulation, it will prevent the C0% from running at 100%. The Task Manager might show you that the CPU is running at 100% but if the CPU is being throttled, the C0% will show a much lower number. It helps to confirm the numbers in the clock modulation columns and makes it difficult for Dell or anyone else to argue against the log file that ThrottleStop creates.

You can try locking the chipset clock modulation at 100.0% and the multiplier at its maximum but you will probably exceed 90 watts DC that your adapter is rated to deliver. It won't catch on fire but it might shut itself off. For gaming you might be OK but I wouldn't be running Prime95+Furmark+ThrottleStop if you only have a 90 watt adapter.

Are you using a Dell? You should be able to call them and get them to send you a 130 watt adapter for free if you find the right person to talk to. You often times have to escalate the problem to someone higher up that understands the throttling issues with their laptops.

If you set your core voltage too low, it's possible that when running a demanding application that the program might become unstable and you could corrupt the OS but that doesn't happen very often. It's a good idea to do plenty of stress testing before deciding on a final core voltage so that it is 100% stable.

What laptop do you have? I might need to add another model to my long list of throttling laptops.

 

24 DTS is something like 80ºC on that cpu no? if cpu shares the same heatpipe as chipset, it will cause the chipset to overheat (they cant handle more than 80ºC-85ºC) and bios has to do something (throttle down). maybe try to open throttlestop after you had your pc shutdown for 1h or something, and check the clock modulation values. if it's diferent than 100%, then say your model to unclewebb
you wont loose anything if you replace the thermal compound and clean the dust from your fans... undervolting is good for lifespawn and power usage, but is not the right way to cooldown a overheating pc. components aren't suposed to overheat if they are in good condictions, if they do, RMA it. I say this because i already opened A LOT of laptops and all of them dont even reach 60ºC after cleaning, but before they used to overheat and shutdown while heavy loaded.
My T9400 reaches 52ºC with 1h of prime95! (fans at low speed) and currently my laptop fan is OFF and DTS is 65! I clean my laptop every 2 months (it's a gaming laptop).
Clean the dust before the dust cleans your cpu

 

Thanks for the thorough explanation (and for the awesome ThrottleStop). Does chipset clock modulation usually just set in once the GPU comes into the mix? It appears to do so here. If I just run some Prime95 number crunching without Furmark it behaves nicely and stays put at 100%. Should this tell me something?

I called support a few days ago and should hopefully receive a 130W adapter soon. I'm running a throttling hall-of-famer from Dell; the Studio XPS 1640 (with the RGBLED screen) which I'm sure you already have on your list.

That's about right. It's somewhere in the upper seventies under maximum load. Further undervolting tests have made this number drop though.

I'll do that thanks. I thought that a natural consequence of undervolting was less power and thus less heat? What am I missing?

 

Of course, less power = less heat... What I was trying to say is that you may lower something like 5 to 15ºC with undervolting but if you clean your laptop you will win anything from 10 to 40ºC less (it depends how much you use the laptop and how old it is). If you have overheating issues 1st clean your pc, if didn't solve, contact your manufacturer, you will have heating problems forever... If you are out of warranty or you are an enthusiast go with undervolting, but make sure that it's stable! I do undervolt my laptop but I don't recomend that as the 1st thing to do when you have heating problems... And dell laptops are problematic with heat but easy to clean. Of course everyone can undervolt a laptop even if it works at normal temperatures, for extra battery and energy savings (and less heat of course).

 

No need to add the XPS 1640 to my list. It's already near the top, especially with the RGBLED. Great screen but that usually sends you over the edge into throttling.

As soon as you start working the CPU and GPU at the same time, it creates the largest power draw and that's when the bios starts using chipset clock modulation to slow things down and to conserve power. It's a pretty dumb design when they thought that a user would not need to run the CPU and GPU at the same time. I guess the engineers have never played a 3D video game.

The 130 watt power supply will help things out but you will likely have to continue to use ThrottleStop to get the most out of your laptop. If you keep complaining to Dell, they might upgrade you to an XPS 1647 with a Dual Core that doesn't throttle as much or at all if you get lucky.

 

Awesome undervolting guide. Unfortunately, I started reading from the beginning and didn't discover that i7's can't be undervolted until - well, now. Oh, well. At least I didn't try it first and post a bunch of "why won't this work???" questions

For the record, I have an HP dv8t, i7 720QM, 4GB RAM, 500 HD, GeForce 230M, 18.4", 1920x1080. If only the keyboard was backlit...

BTW, Dell's been having throttling/overheating problems on all their i7's. I started with a Studio 1747 and returned it because it had more issues than DC Comics.

 

Thank you Falesi! In anticipation of my new system I stumbled upon this thread and bookmarked it in hopes of applying this for when the system arrives. Luckily I jumped to the last page to see what people are saying about this technique and saw yours before I attempted this, as I also have a i7. Thank you again for posting!

 

Just wondering but are there a similar guide to undervolt GPUs?

Would be cool if I could undervolt both the GPU and CPU

 

I use nibitor to play with my GPU (nvidia) and it is possible to undervolt it if you are lucky. You only have 2 or 3 VID's avaiable (that cannot be changed - always the same voltage - not controled by bios). To me, undervolting was possible with underclocking a bit, so... i decided to overclock it almost 30% with stock voltage (yeah... i dont want to loose performance)
For mobile gpu's, tweaking the bios is the only way that i know to undervolt...

 

When you say "BIOS", do you mean the VBIOS (or in other words the graphics driver), or the actual system BIOS? If it's the latter then I really wouldn't go screwing around with it, bricking the BIOS is about th only thing that can truely brick a computer for good other than just breaking it.

I'm guessing the reason they only have three voltages is for the three different profiles, one for the 2d profile, one for 3d and one for the throttling profile.

 

With my laptop I realy mean BIOS because the VBIOS is a module inside the asus g71v bios. I use MMTOOL to extract the module, nibitor to tweak, mmtool again to replace the module and winflash to update the bios.
But most of you have a "normal" vbios. Check if gpu-z can read your bios. If it can't is probably because it's a module inside your laptop bios.

I only have 2 voltages. 0.86v to 3 profiles and 1.16v to the 3d performance profile. There is no way to change the 1.16v to 1.0v for example. But I can change the 3d performance profile to run at 0.86v. If it's unusable, no problem, your laptop will crash, just flash with another bios. If you don't change the boot up clocks, there is almost no risk to damage your card (make it unbootable).

 

Dammit, I think I may be in that same categroy then. GPU-Z will read my BIOS but it won't extract it, which is what I want it to do as I want to change the 2D profile.

As for voltages, there are voltmods out there.

 

If you have an AMI bios you can try to open it with MMTOOL and see if there is one module called "pci option rom" with the same ID has your graphics card... Since you have an ati card, nibitor won't help you but maybe you can try to extract your bios with that. Or use one ati bios editor to extract... i never used one so i don't know how to play with it, sorry.

edit: try atiwinflash

 

I was planning on using RBE to modify the 2D profile so it had pretty much no power. I was trying to use the atiwinflash thing too and it gave me an error message, "ReadBIOSToArray".

I'm not the most in the know about GPU BIOS flashing so I'm still learning, there are new terms and stuff I'm not the most farmiliar with. I understand the concept of changing clocks and undervolting, I've done it several times before but the idea of having to start up a clocking program every time I start windows doesn't appeal to me so I thought I might do it properly.

 

I have undervolted my T5670 to just .9500v from 1.1750. Temps have reduced from 85 to just 70c. I ran stress test for 1hr and played dirt 2 for half an hour.
I really thank for this wonderfull guide

 

Hi, just got my m1530's mobo replaced with a refurb one after so many difficulties, so i don't want my computer to die again for overheating problem.
The problem i am facing is that i am not able to understand if i have to lower the voltage of every multiplier, if yes do you have any idea on how much i can lower in a t9300 processor.
Thank you

 

With a T9300 i recomend you to try throttlestop. Just select the maximum multiplier and lower the voltage while you have prime95 opened, running with small fft's. Once you got an error, increase the voltage by 0.1v and you are done.
Remember that T9300 has IDA (aka turbo mode) and your system might be stable with 2 running threads but 1 thread it may crash (because of ida.. extra speed). So test it with 1 thread instead of 2.

A laptop must be opened and cleaned at least 1 time a year or else it will start overheating because of dust. I clean mine every 2 months
consider undervolting as an alternative and not a solution to overheating.

 

Thanks TigTex, but i am not good with software's so i am not able to understand "throttlestop", "prime95", "small fft's" and about 2 threads and one thread.Please make me clear for this.
Thank you.

 

sorry for bumping the post, but can anyone help me asap as from tomorrow i am going out of station and will be unable to check the solution, so guys please help me.
thank's

 

my T9300 setting for your reference:



Note:
- uncheck the "Engage Intel Dynamic Acceleration (IDA)" in Advanced CPU setting

 

Love you crayonyes thanks a lot, but will this be stable in my computer?

 

That's why you'll still need to test the stability
if it's not stable put a higher VID and test again, and so on
But this setting is rock stable in my Vostro

 

Can I applied this guide for i7 Quad lappy?

 

hey crayonyes should i worry about what TigTex was saying to do?
By the way rep+ for you

 

no buddy i don't think you can undervolt your notebook using this guide@udomwattawee.
See post by dimmok in page no. 503

 

by checking IDA in Main profile & performance on demand,
and unchecking the Engage IDA in Advanced cpu setting,
the T9300 will run normally @2.5GHz, so no problem with RMClock.

 

Hey in my profile page it's not showing that low vid for each multipliers as yours, so how can i have all the vid's same as yours.
Thanks for your prompt response.
EDIT:-is it needed to lower vid for every multiplier?

 

yeah, they're not always the same for each notebook.
That's why it's only good for reference. 0.925V is my lowest VID.

Yes, lower the VID on every multiplier you're planning to use

 

one last doubt, flipfire says to untick superLFM and IDA, so should i tick only IDA as per you or both.
Thanks again.
EDIT:-If i will uninstall rmclock, should the settings revert back to the original factory settings?
What shall i do to get factory settings back in case i feel the notebook is not performing well.

 

superLFM is for super battery saver mode when in idle, it only runs @ half-FSB
and for T9300 which has half-multiplier, tick IDA in profile and untick Engage IDA in advanced setting will make it runs 2.5GHz, not 2.4Ghz.

The choices to use them or not is up to you

Closing (Exit) the RMClock is good enough to make the setting back to original.
or you can run the RMClock_WipeOut.reg in C:\Program Files\RMClock

 

hey thanks crayonyes i have underlocked and my cpu ir running 12C cooler.
One last and final doubt, as i have underlocked quite a bit so i experience some down performance, but the settings are in high performance, so my doubt is how to use that settings in power saver option i.e how to underlock in power saver option.
thaaaaaannnnnnnxxxxxxxx alot.

 

underlock << you meant undervolt right ?

I'm not sure I get the point of your last question,,
but if you experience any lag, check if you had tick all the FID index on the performance on demand profile.
And are you using windows 7 ? If yes you can change directly to power saver scheme in windows power options
You'll still have it undervolted, cmiiw -check the voltage with cpuz --

 

Just used this guide on my XPS and might I say thanks to the OP for putting this together!

My CPU was 87deg C now it's 53deg C
My GPU was 80deg C now it's 72deg C

So a great improvement for me.
I want this to run at all times (in the background without opening it all the time) what settings do I use under;

Management > CPU default settings?
At a guess I would uncheck;
Restore CPU defaults on management turning off &
Restore CPU defaults on application exit

But what about, CPU defaults selection?

 

oh, yes undervolt.
By the way i mean as i have undervolted my notebook, but some times i need high performance, at that time how can i use the original factory settings without changing the vid and all that .
Thanks.
EDIT:-As per flipfire when we are done with the undevolting we have to change the startup seetings under profile to performance on deman, so we have to do it for both battery and ac power or for only battery?

 

Unless I'm missing something...
That's why you run a stress test so although you have undervolted it, it will still run high performance.

 

Undervolting doesn't reduce performance... it reduces stability if you don't test it enough.

But if you disable IDA, you will loose performance... That's why I was recomending "throttlestop".
ThrottleStop 2.54
http://www.mediafire.com/?mh1wn2njmg4

It's easier to configure and produces better results than rmclock, without any impact in performance.

 

But i don't know the procedure to be followed on throttlesstop.

 

hey crayonyes please help me as i will not be able to check the solution tommorow.