The "Undervolting" Guide

You still haven't said what CPU you are trying to use RM Clock on. Is it a secret?

I didn't pay for RM Clock and Profile section that you are showing is full of values that I can adjust. You don't need the PRO version to access this basic feature.



If your CPU was supported, it would show the same.

 

OK. Sorry. I'm using the LG X130 Netbook with Atom N280 Processor.

 

RM Clock does not support the Atom CPUs as far as I know.

When Intel Atom CPU N270 will support? - RightMark Forums

 

I might be a little late in adding this but i made a video tutorial of this undervolting guide. It's cut down alot but I've credited flipfire in the video and there's a link to this guide in the description. As far as I know no one else has made a video, so I thought I'd do it first

YouTube - How To Fix Throttling

And please, no insults on my Scottish accent .

 

Hey, nice post !
I tried the ORTHOS cpu loader and the HWmonitor, the min and max temp of my cpu (max 54 load for 10 min), Then i change my i change the multiplier to 1.100, my cpu temp is still the same but there been a change in my gpu temp from 72 c to 66 c , thx...
by the way, im new here..

 

hey, does intel core 2 duo processor supported? it only has 1-4 index...

 

what type of intel core 2 duo u got?

 

My Laptop "aspire 5920g":
Intel Core 2 Duo T7300 2,0 GHz
I managed to lower my volt by 0,200v (1,2375v-1,0375v) without getting any bluescreen yet.
my max temp was 89c, now it 68c max. thats 21c shaved off!. thanks flipfire for your wonderful guide.
without you, I wouldnt even know what undervolting is.

 

Does anyone know if the Core i5 processors are support in RM Clock or Throttlestop? If not is there something that does support them?

 

Development ended for RM Clock long before the new Core i processors were released so RMC does not support them.

ThrottleStop supports the Core i5 but Intel has eliminated the ability to easily change core voltage like you could with Core 2 mobile CPUs so ThrottleStop is not going to be able to do a lot for your CPU other than monitoring its core temperature and performance.

 

I did some testing with wPrime and Clock Modulation and turned down my modulation to 75%. I thought that this was keeping my laptop a little cooler but maybe it's just placebo.

Would clock modulation help?

 

Clock modulation slows your CPU down internally. A ThrottleStop setting of 75% is putting your CPU to sleep 25% of the time. A CPU that is sleeping will consume less power and run cooler but obviously it won't be able to perform as fast. If you want a slow computer with long battery life then maybe a netbook might be for you.

Seriously, the best way to save power is to let the CPU run as fast as possible so it can get its work done as quickly as possible and then get back into one of the low power C3/C4/C6 sleep states where the real power savings are. In the Core i design, when a core enters the C6 sleep state, the voltage is reduced to zero. It's pretty hard to improve upon that.

 

Interesting... I have a P8600. with your above statement, does it mean that I should leave the "Power Saver" feature unticked in Throttlestop if I want to conserve battery?

 

flipfire, thx for the epic guide man !
I never got a BSOD and managed to goto the lowest VID
If u guys are interested here are my results:

 

Vodzilla: Here's the testing I did to try to get to the bottom of whether high or low FID / VID combinations are good or bad.

http://forum.notebookreview.com/6369641-post5060.html

With a 45nm Core 2 mobile CPU, it doesn't seem to make any difference to power consumption at idle no matter how you have them set up. When it comes to power savings, allowing a CPU to run as fast as possible doesn't seem to make sense. It doesn't make sense to run a car as fast as possible but CPUs are completely different. They are most efficient when running at full speed. This allows them to get their work done quickly and to spend more time in the low power sleep states where the real energy savings are.

Do your own testing. The Power Saver feature was designed for the 65nm Core 2 Duos but doesn't change power consumption any on the newer 45nm Core 2 Duos.

 

ok cool, I see what you are saying. Not intersted in a netbook though. I may just roll back to my E4300 as it runs much cooler and all I do is surf the Internet and use Office occasionally.

 

@unclewebb - wow that's enlightening. thanks!

 

The 45nm Core 2 Duo CPUs are designed to automatically go into deeper sleep states like C3/C4/C6 at idle and can spend as much as 99% of their time in one of these states. The internal VID voltage for C3/C4/C6 is lower than the VID setting that ThrottleStop or RM Clock lets you access. That's why in my example above, it didn't matter whether I used high or low FID/VID settings because these settings are barely used. At idle, these settings that might have been important for saving power and reducing heat in the 65nm Core 2 mobiles CPUs, no longer applies to the newer CPUs. The only important thing to do is to minimize your maximum VID and that's about it. These CPUs take care of themselves after that.

There's no harm in using RM Clock and setting up each multiplier and VID individually on the 45nm CPUs but these processors are designed to transition the multiplier very rapidly from the minimum to the maximum. SpeedStep can make multiplier changes in as little as 2 microseconds which is way too fast for any monitoring software to report accurately. Most CPUs are going to spend close to 0% of their time at the intermediate multipliers so there is virtually no power to be saved by getting overly technical and testing for hours to set these perfectly.

It's easy enough to use the Windows Performance Monitor to see what percentage of time your CPU spends in C3 or a deeper sleep state. If your computer is set up properly and your CPU is spending most of its time in C3 or deeeper then it doesn't much matter how your idle VID is set because it's not being used anyhow. When you apply a load, it's going to transition up to the higher multiplier and VID value almost instantaneously. Holding one of these CPUs back at the 6 multiplier for any length of time makes it slow and inefficient but doesn't save you anything.

essense: The VID numbers you drew a rectangle around in RM Clock are your VID requests. Make sure the RMC monitoring graph shows that you are actually getting that voltage. There's an obscure setting in RMC that controls this when you request less than 1.00 volts on most CPUs. On the Advanced CPU Settings tab, you usually need to have Enable Dynamic FSB frequency switching enabled to get access to the lowest possible voltages.

 

Thanks unclewebb, I have confirmed that the voltage stays at 0.950 and the "Enable Dynamic FSB frequency" box seems to be blocked for me

But i'm happy with these results

 

unclewebb dont you have a guide of how to use your program? first time that i will do undervolting, i have a dell xps m1530 with core2duo T7250.

i download it but now i dont know what to do

thx for ur help

 

Will RMClock work with Windows 7 x64?

 

If you had read the first post of this thread, you would have known it will work on 64-bit Windows. And there are enough posts from people mentioning Windows 7, which would cover the rest of your question.

In short - yes.

 

Push the F1 key when ThrottleStop is open and running and you can learn some of the basics in the help file. ThrottleStop is designed for people that aren't afraid to click on some boxes and to see how their choices affect their laptop's CPU and battery performance.

Here's the short version: Download ThrottleStop, start program, click on the Turn On button, put a check mark in the Set Multiplier box, reduce the Voltage ID box until your CPU loses Prime95 or similar full load stability. Once you have found a stable voltage then you are pretty much done. With a T7250 you can also try the Power Saver option. If that drops your idle core temperature then that's a good thing but on many of the newer 45nm CPUs, it's not worth bothering with. I don't have access to enough hardware to know what features will work best for you so your best is to play around with ThrottleStop and try to find out what it is capable of on your CPU.

If you have any specific questions then send me a PM message. Make sure you read the docs first and play around with the program for a few days before asking questions. RM Clock should work great on a T7250 so if you need a manual then just head to the first post in this thread.

 

Hi! I am using ThrottleStop to undervolt the SL9400 in my 2530p. The multiplier I should force and adjust is the 1.86ghz one (7) or the IDA one (8)? Thank you, I hope is not a repeted question but I searched the thread and could´nt find something related.

I suppose is the normal one (7 to get 1.86ghz) but I want to confirm. I think the IDA voltage must not be forced. Should I use RMclock to adjust both (IDA and the normal one) ? I kind of like throttle stop better.

My battery life will be the same regardless of the multiplier if my voltage is equal for both?
Example: 0.925v for 7 (1.86ghz) and 6 (1.6ghz) Would'nt this make the LV chips a lie? I can put my p8600 with the same multiplier and voltage from my SL9400. Finally does the UKV (su9400) have lower voltages than 0.925 that can be set. Wanted to change my 2530p for the su9400 version but I am having second thoughts. Sorry for the long post.

 

Here's my results with using the first post of this thread:

HP 8710P w/upgraded T9300 from a T7500

I had no idea if I should be using IDA because I saw several conflicting posts about it, so I left it and option 8 enabled.

Defaults:
6, 7 and 8 were set at 1.375V and max temp was 67c

Modded settings:
8 | IDA | 13.0x | 1.0500V | max temp 66c
7 | Normal | 12.0x | 1.0250V | max temp 60c
6 | Normal | 11.0x | 0.9500V | max temp 55c
5 | Normal | 10.0x | 0.9500V | max temp 52c
4 | Normal | 9.0x | 0.9500V | max temp 52c
3 | Normal | 8.0x | 0.9500V | max temp 52c
2 | Normal | 7.0x | 0.9500V | max temp 52c
1 | Normal | 6.0x | 0.9500V | max temp 52c
0 = Disabled

For settings 6 - 1 RM's voltage settings couldn't go low enough for my system... strange.

CPU idles at 26c

 

Totally cool guide! I used your guide and I undervolted my core 2 x9100 from 1.2375v to 1.0500v. I tested it as you said and I have no blue screen. How much did I gain in watts? I got like 30c less under full load! pretty cool! And it also keeps my sli graphics cards cooler since my system is cooler overall. Is there a way to undervold graphics cards without losing performance? Btw for my processor I get the lowest as 1.0500v. Can I get it less? Thanks in advance!

 

First, I'd like to thank flipfire for writing this guide and all the others (theres 500+ pages... no way I'm going through everyone! haha) who have contributed to the thread allowing such easiness in undervolting, because without you guys my laptop has never been cooler.

But, now I have a problem. Originally, I ran a stress test and it went up to 95 degrees (hot!). After undervolting successfully, the stress test showed the cpu to be around 70 degrees. Major difference.

I saved my settings and made it run during startup as shown. This was about a month ago, and everything was running smoothly... but recently I noticed my CPU getting quite hot again. So, I ran the stress test again with the undervolt in RMClock.. and for some reason it's right back up at 95 degrees. I double checked my voltages and even monitored "CPU Info" under RMClock, and all the voltages were the undervolted ones...

So now I'm confused on why the voltages seem to have decreased, but the temperature is so high again. I've been trying ThrottleStop without luck either.. anyone have any ideas?

 

If you reinstate the default voltages do the temps go higher than 95?

 

Just open your laptop, clean the fans and replace the thermal compound.
Probably your pc is full of dust and the fan can't create an reasonable airflow.

 

^ true, and it did wonders for mine.

 

Uncle Webb, I have a Core i5-540M and the SLFM check box is grayed out on my Dell Studio XPS 16.

Any ideas?

UPDATE:

A sad day for undervolting

 

I trying to undervolt a SL9600 in a x200t but it seems as though the lowest VID available at any multiplier is 1.0375v. Is there anyway to unlock lower VID or do I possess a poorly binned old top end chip?

thanks

 

kb8to24: If you are using RMClock, go to the Advanced CPU Settings tab and make sure Enable Dynamic FSB Frequency Switching (DFFS) is checked. You sometimes need to do this to access the lowest voltages. If using ThrottleStop, make sure SLFM is checked.

 

does this work with core i7s as well or is it a different guide?

 

Core i's can't be undervolted like this. Currently no one knows how to do that.

 

Maybe the "Force TDP/TDC" option helps if the only thing you want is more battery life at the expense of performance. Has anyone here lowered their default TDP with ThrottleStop? If the Core iX processors are smart enough they will do the rest for us.

 

The TDP/TDC setting in ThrottleStop controls the amount of turbo boost your CPU gets. Reducing the amount of turbo boost will slow down your computer but it might not save you any power. At idle, individual threads in the Core i are designed to enter the C6 sleep state where the core voltage drops to zero. Using turbo boost will cause your CPU to run faster, get tasks done quicker and spend a higher percentage of idle time in the C6 state where it is consuming virtually no power. It's hard to improve upon that.

To test this, it would be easier to select Disable Turbo and get rid of all turbo boost if you think this could be a power saving feature.

In the 45nm Core 2, I've found that a slow CPU is an inefficient CPU so going slower does not necessarily save any power. A slow CPU reduces the amount of time the CPU spends in one of the low power states like C3/C4/C6 where the real savings are.

 

yeah, its checked but the lowest voltage is still 1.0375.

shouldn't sl9600s typically be binned better than sl9400?

 

And I was just about to change my sl9400 for a su9400 Elitebook 2530p. I suppose the 0.875v, 1.4 ghz isnt much of a difference to the 0.925v, 1.86ghz setup I have at the moment and I get to have 6mb of cache.

Unclewebb which setup do you recommend? I am not sure if the minimum undervolting voltage for the su9X00 is .875v

 

To be honest, the minimum voltage is mostly a meaningless number in the 45nm Core 2 CPUs. Your CPU rarely uses this value at idle. If set up properly it will go into one of the C3/C4/C6 sleep states and spend 99% of its idle time there where it uses a completely different and lower VID anyhow.

Extra cache is usually a good thing.

 

Agreed and also undervolting at your load clocks is where you will see big temp and power reductions.

 

Here's what the Intel spec sheet shows for the SL9600:

Code:

 C4 : VID 0.650 - 0.850
DC4 : VID 0.600 - 0.850
 C6 : VID 0.350 - 0.700

You can use the Windows Performance Monitor to make sure your CPU is going down into the deeper sleep states. At idle, this is where the power savings are and most of these newer CPUs do this automatically regardless of any settings in any software like ThrottleStop or RM Clock.

 

Undervolting arrandale (i3/i5/i7) with software is not possible yet.
Also for many other CPUs minimal voltages are disabled by Intel.
I am interested in the true minimal voltage of my new 32nm-i7-640m. (!)

In 2008 I did some research about minimal voltages on differnt Notebook CPUs with rmclock.


In 2009 I did some reseach on optimizing thermal compounds for notebooks.
Quiet and Cool. Thermal paste replaced! 38@idle 73@load

And I did some basic research on voltage pinmods.
Core2Duo T5250 (and the same) hardware voltmod(undervolting) !? [1] - RightMark Forums

Now I found additional information on an Intel VID override Circuit.

Source: http://download.intel.com/design/chipsets/embedded/323094.pdf

The Solution: The "official" Intel "VID override circuit" may help for an propper pinmod with respect to the signal-IOs
and by the way avoiding the risk of shortcutting the signal drivers.

From Intel-datasheet (i7=Arrandale 32nm):
VID[5:3] and VID[2:0] are bidirectional.
As an input, they are CSC[2:0] and MSID[2:0] respectively.“
Source: download.intel.com/design/processor/datashts/322812.pdf
So these inputs combined with the pinmod can affect the boot-sequence by giving wrong signals to the CPU
(at least at i7-Arrandales, but probably also for Penryns etc.).

CSC[2:0]/VID[5:3] - Current Sense
Configuration bits, for ISENSE gain setting.
This value is latched on the rising edge of
VTTPWRGOOD.
MSID[2:0]/VID[2:0]- Market Segment
Identification is used to indicate the
maximum platform capability to the
processor. A processor will only boot if the
MSID[2:0] pins are strapped to the
appropriate setting (or higher) on the
platform (see “Market Segment Selection
Truth Table for MSID[2:0]” on page 88
for MSID encodings). MSID is used to help
protect the platform by preventing a higher
power processor from booting in a platform
designed for lower power processors.
MSID[2:0] are latched on the rising edge of
VTTPWRGOOD.

Several of the VID signals (VID[5:3]/CSC[2:0] and VID[2:0]/MSID[2:0]) serve a dual
purpose and are sampled during reset. Refer to the signal description table in
Chapter 6 for more information.

Maybe a small circuit can be build for flexible overriding the VID of actual Intel CPUs
e.g. by a combination with a flash EEPROM?
VID-Input to Address lines and Data lines to VID-output.

 

Penryns just use VID as output only, no configuration pins. The above circuit is just a switch for testing at fixed voltage and probably just used on evaluation boards for testing. Voltage is set by jumpers as is the actual switch from VID control to jumper control.

Trying to place a small circuit between the CPU VID lines and the VRM would certainly IMO be ambitious. AFAIK SIO chips have long had the ability to change VID line encoding but it's up to the manufacturer to incorporate it in hardware.

Strange that Intel disabled VID control via register.

 

Yes, but this minimum voltage is available for the maximum multiplier so when loaded the cpu will use this voltage. Atm I am truly happy with my setup (0.925v, 1.86ghz), undervolting did wonders for the noise and temps in my Elitebook (excellent notebook except for the noise).

 

That's absolutely true. I forgot about that because on my Core 2 Duo, I can't run an SLFM voltage reliably with the highest multiplier.
Sounds like you've got a good CPU.

 

I have a Intel core 2 quad q9000 and before undervolting i was getting low 90s and I tried to undervolt it but I'm not able to make the multiplyers any lower and the lowest voltages I can select are .1v less. After undervolting I get high 80s but I was wondering if I have any other options for undervolting further.

 

Dimmuklok: What software are you using? Post some screen shots at full load if you can. Upload them to ImageShack® or similar and then post a link to the thumb nail image here.

Does the voltage not go any lower than 0.1v less or is that as low as you can go and still be stable?

If using RM Clock show some graphs or use CPU-Z or HWiNFO32 to show some info about your CPU.

 

Ill post pictures in my next reply. If you go to page 11 of this thread and read about falcon2claw's problem its pretty much identical to mine. The problem was that his cpu was "high performance" and because of that RM Clock would not allow him to go below .1v or x6 multiplyer, identical to what is happening to me. The conclusion they came to was that he basically was screwed and he had no options so I was posting if we have found anything out since then.

I use CPUID hardware monitor and I ran ORTHOS to make my cpu work hard. I got low 90*c before undervolting and I only ran ORTHOS for a couple mins but it seemed like it wont go any higher than that. After undervolting it was high 80s*c.

RM Clock will not allow me to go below .1v or x6 multiplyer. When you click the voltages in RM Clock you get the drop down menu and you can pick what you want to change the voltages to and for me it only ranges from 1.265 to 1.165. Its completely stable at 1.165 I just don't the option to lower it with RM Clock, and if there is the option I don't know where it is.

Thanks for the help Unclewebb. Pictures are still coming.

 

I don't know if I'm doing this right; I haven't been shutting down and letting it cool down after;

I have a T7500 @ 2.2 GHz

I lowered the 11x mult from 1.325 to 1.2875 and I did not see my load temps drop really. Like 1 C difference. It's 70C full load.