The "Undervolting" Guide

Or... try throttlestop and enjoy an extreme easy and efficient undervolting
just an idea... Your cpu will work better with throttlestop, and maybe you can unlock the dualIDA mode (more speed). Don't see a reason for not trying throttlestop in your case, since rmclock isn't even working like it should

 

Here's the problem with how RM Clock reports Intel Dynamic Acceleration mode.



It shows a nice flat yellow line indicating the 10.0 multiplier and a nice flat red line indicating the 10.5 multiplier even though core 1 is using the 11.5 multiplier over 95% of the time. RMC gives you no indication if IDA mode is working correctly or not. There are a couple of purple blips downward but that's just monitoring error.

 

is it necessary to set the Minimum processor state to 100% when using ThrottleStop, or does this only apply to RMClock?

and if i want RMClock to autostart when windows boots, is it better to set it from Windows Task Scheduler, or should i just simply checkmark the "Run at Windows startup" box in the RMClock settings? if i should just check the box, then should i select Startup by registry key or Startup folder?

 

You need to do that for both programs so they can have control of your CPU without having to fight with the Windows CPU power management schemes. For RM Clock, setting this at 100% is very important in Windows 7.

Use whatever method you like to get RMC to start up with Windows.

 

unclewebb, i just wanted to give you a bunch of thanks and kudos for keeping this thread alive and being the main person to help/answer all questions in here. big thanks for the programs you created also. i use realtemp on all my intel computers.

now to pester you with another small problem, have a CULV processor that only has the 6.5 multiplier available, and i did the small registry hack to enable that half multiplier already. i used rmclock to decrease the vcore to the lowest possible (0.875v), and it looks like it's prime stable. however, when i just use the notebook normally to browse the internet, it'll crash within 5 minutes. for some weird reason, it seems to be more stable when under load.

any reason to this? i'm thinking it's because there's some other power saving feater conflict between the OS or some other device.

 

Your welcome WhiteFireDragon.

I had a look through the Intel Specifications and found that when a CPU goes into the C6 sleep state, the voltage can drop down into the 0.35 - 0.70 volt range. It goes lower than the VID settings that are available in RM Clock or ThrottleStop.

If your CPU is not stable at this low voltage then its possible that when it is lightly loaded, it is going into one of the deeper sleep states and locking up because of a lack of voltage.

In RM Clock in the Advanced CPU Settings, my T8100 has an option in the Chipset section called enable C5 and C6 modes. That's checked on my computer by default so if you un-check that, maybe it can prevent your CPU trying to enter these states which might help with the lock up problems when lightly loaded. There shouldn't be a significant different in power consumption whether this is enabled or disabled.

You might want to try some LinX testing. Ordinarily I don't recommend this for laptops because of the amount of heat it creates but your laptop runs at a low speed and low voltage so it shouldn't get too hot. A 10 or 15 minute test might give you a better idea if your CPU really is 100% stable at 0.875 V.

LinX - A simple Linpack interface - XtremeSystems Forums

Also do both Prime95 Small FFTs and some Prime95 Blend testing. Hopefully you can figure something out.

 

Undervolted my Toshiba laptop a couple months ago using this guide (thanks, flipfire!), and it has been running much cooler.

But now I'd like to undervolt my little Asus netbook with an Atom N270 processor. RM Clock won't work for this, so I was wondering if anyone could point me to an app or program that will do so without me re-reading every post in the thread to see if the answer is hidden there somewhere. Thanks for your help!!

P.S. My netbook already has something called "Super Hybrid Engine", but I'm assuming all that does is underclock the processor (depending upon the setting), rather than undervolting it.

 

An Atom cpu has a very low TDP (thermal design power) so undervolting him is useless. You can do it, but the results won't be interesting at all. Maybe you can save 0.2w (5min on battery?...). Your netbook's chipset uses more power than your cpu so 50-60% of the heat is caused by the chipset.

 

Fair enough. I much appreciate the info, TigTex. I'll just leave it as is.

 

(I'm a Windows 7 64bit user)

You're a life savor.

I was running at 105C (MAX load) on my core duo 7500, and now I'm running at 80C (MAX load) by lowering it from 2.2500V to 2.1000V.

It was as bad as frequently getting CRC errors when downloading and extracting large files, and I passed all the memtest86 and sector hard drive scan as well. I can attempt to set the voltage even lower, but I think I'll be OK as long as things continue to work properly.

I'm running at an idle at an idle 50C now(I used the auto adjust option to adjust the voltage), but it could be cooler if I let it sit idle longer.

Many thanks.

 

with 2.25v your cpu was already burned... 1.25 maybe no?

Those temps are still too high. You MUST open your laptop and clean the dust and replace thermal compound. 105ºC is the tjmax of your cpu, you may have damaged your cpu by constantly hitting that temperature causing crc errors... be careful next time, clean your laptop every 6 months MAX. IF it keeps giving you errors and bluescreens, you probably destroied your cpu...

My cpu it's 50ºC while playing games and max 66ºC with inter burn test... T9400

 

The T7500 is 65nm so it's going to put out a lot more heat than a 45nm T9400. You're right that 2.25 volts would send it to an early grave.

 

In its MSI GX623 (C2D T6600 Penryn) I have a problem with the CPU Whining.
I read that a solution can be RMClock program. Unfortunately, it does not recognize my chipset thus can not exclude, for example, C4 mode.

photo:
Imageshack - schowek02ad.jpg


Is someone able to help?.

Could You give me what I should set the voltage to eliminate the whine?. I have C2D T6600.

 

Try using ThrottleStop and toggle the C States option on and off. It probably won't help that problem but it never hurts to try. You can also toggle Power Saver and SLFM to see if that makes any difference.

Every CPU is unique. The only way to find out how much voltage your CPU needs to be stable is to run a program like Prime95 and lower your voltage until you lose stability.

 

well i am trying to undervolt p8400 but i have no clue of what is happening.i am using throttlestop.while prime95 running in the background cpu-z is only showing change from 1.000v to 1.025v.if i lower voltage from 1.000v then cpu-z remain stuck at 1.000v and also if i set voltage higher than 1.025v,the cpu-z continually show 1.025v.below are some pics




infact i lowered voltage to around 0.9500,but i saw no change in cpu-z.
plz tell me what i am doing wrong.also can anyone plz tell me step by step process of undervolting using throttlestop.

 

I couldn't see the pictures.

Try using ImageShack® - Online Media Hosting

You have to select the SLFM option in ThrottleStop if you want the lowest possible voltages. Did you set your Minimum processor state in the Control Panel -> Power Options to 100%?

The best way to under volt is to run a load like Prime95 Small FFTs and then keep lowering the core voltage until you get an error in Prime95. Your computer might also get a BSOD when doing this and lock up. That's a sign that you've gone too far. I also use HyperPI or wPrime. If you are getting errors in those programs then you need more voltage.

Edit: Your links weren't showing up but I got the second and third ones working. Anyhow, on my T8100, if SLFM is not enabled, the lowest voltage I can access is 1.000 volts. With SLFM enabled, then I can go lower. The lowest this CPU allows is 0.9500 volts.



When testing also try running a single thread of Prime95. This will let your CPU use Intel Dynamic Acceleration mode where the multiplier on the core doing the work will cycle between 8.5 and 9.0 continuously. Make sure that is stable too.

 

thanks unclewebb for help.i accidently put wrong links for the pics,but anyhow you understood the problem.i have set minimum processor state to 100% in power options.
i have just one question though,will there be any negative effect on the performance by enabling SLFM?.Also different versions of cpu-z gives different voltage readings?...I am now using version 1.55 which is quite accurate.
how can i run single theaded benchmark in prime95,and whar priority should be used?
i will follow your tips and will post my results.once again thanks.

 

Select Torture Test in the Options... menu and then this screen should pop up where you can select if you want to run 1 or 2 threads of Prime95. By selecting 1 thread, that will allow your CPU to use Intel Dynamic Acceleration on one core, probably about 90% of the time, while the second core spends most of its time in the C3/C6 sleep state. This will show you if the voltage you've chosen is OK for all types of use, full load or partial load. I just use the default or normal priority that it gives you.

Download Prime95

I think the latest version of CPU-Z should be reporting the VID voltage correctly. If you're not sure then you can also try HWiNFO32.

HWiNFO & HWiNFO32 - Hardware Information and Analysis Tools

That program will give you a complete table of the VID values for your CPU for SLFM, LFM, HFM and IDA mode.

I've tried some benchmarks before with SLFM enabled in ThrottleStop and it didn't seem to make any difference that I could measure. You can try running 3DMark06 with and without this being enabled to see if there is any difference. When I get the chance I'll probably go run some Everest benchmarks just to see.

With the Minimum processor state set to 100%, if you want a low multiplier and VID at idle then you can try using the ThrottleStop Power Saver feature. I find that it doesn't really save you any power at idle but some users like to see a low FID/VID at idle. I find these settings are mostly ignored by the 45nm CPUs at idle because they are actually using lower voltage values than these internally that software can't adjust.

 

thanks for amazing fast response.i am running prime95 at vid 0.9375 and its running good at the moment.will lower it further offcourse.temps droped from
90C at 1.025v to 73C at 0.9375,so its working quite good.
thanks unclewebb for this amazing application.

 

well in first testing phase running prime95 on two cores,i got BSOD at 0.9150v while being stable at 0.9250v.
in second testing phase running prime95 on one core only kicked in IDA mode at 9.0 multi and i got BSOD at 0.9350v.
so i settled for 0.9500v for i got no BSOD.
things were looking good here.
now as soon as i start throttlestop with 0.9500v at idle,the VID varies from 0.9500v to 1.000v in cpu-z and after few seconds pc hangs and i get BSOD.if i increase VID to 0.9650 then i get no BSOD.
my question is why i get BSOD with 0.9500v at idle while it remains stable with prime95?
also why it is varying between .9650v to 1.000v in cpu-z although i have 0.9650v in throttlestop?
just a thing to tell you unclewebb that i had to manually select affinity for one core in task manager for prime 95 process after selecting torture test for one core in prime95 as only then IDA mode kicked in.

 

Finally settled for 0.9625v full stable.ran wprime and 3dmark06 and got no errors.temps have droped from 90C to 75C in 3dmark06 which is amazing none the less.And apparently there is no performance drop by enabling SLFM.
thanks for all the help unclewebb,you are the man..

 

Now that you've had fun playing with ThrottleStop, do you want to try doing the same thing with RMClock? RMC lets you set one voltage for your default multiplier and a higher voltage for the IDA multiplier. That should let you use a lower voltage for your default multiplier without getting any BSODs.

Now that you have a good idea of how much voltage your CPU needs to be stable, it would be very easy to set up RM Clock and give that program a try. It takes a little more time to set up compared to ThrottleStop and you'll also need to do a registry trick to fully support your CPU but you might get slightly better temperatures.

Did you set the Minimum processor state to 100%? Usually if that is set then the VID as reported by CPU-Z will be stable. Try HWiNFO32 if this is set to 100%.

If you don't use Set Affinity..., IDA mode will still work but it will be harder to see because it will be rapidly switching back and forth between the two cores. If you open up two instances of CPU-Z and right click on it so you can monitor each core, you might be able to see this happening. ThrottleStop should show the average multiplier bouncing back and forth between the default multiplier and the IDA multiplier, especially if you check the More Data option.

If you are interested in trying RMC and need some help setting it up correctly, just let me know.

 

how do u undervolt using throttlestop unclewebb?

 

If you have a Core 2 CPU, download ThrottleStop, start it, click on the Turn On button so it shows ThrottleStop Enabled at the top right, put a check mark in the Set Multiplier box and then for the Voltage ID just below that, click on the down arrow and reduce your voltage. After that run some stability tests to make sure your laptop is stable.

As mentioned above, for some users RM Clock offers more control and might be a better solution if you are willing to play with it. For other users like myself, having perfect voltage settings for each multiplier isn't that important. I usually lock my laptop to a single multiplier so having a single voltage adjustment is good enough for me.

 

its not possible to undervolt core i5-520m???

 

yes currently all intel core i series cpu's cannot be undervolted.

 

I don't think it's a case of "currently", apparently they actually took out the feature which lets you set the VID. However, I'm hearing about pin mods which change the voltage but they're very limited in what they can do.

 

A late reply, sorry about that

I have a Intel Mobile Core 2 Duo P8700 in a Dell Vostro.

My goal is to make this little thing run less hot!

I used rmclock before but it stopped working as I wanted it to.

If i turn on throttlestop and check the multiplier box and then lower the voltage, the laptop will always run on multiplier x10 (which is max on my cpu), is that correct? I suppose it's called throttling when the cpu changes the multiplier?

What downsides does throttling have?

As i said, my main goal is to undervolt my laptop to make it run cooler while still having maximum performance.

Does is affect batterylife when not throttling?

A lot of questions and i'm sure i sound confusing

 

A P8700 has a default multiplier of 9.5 and when Intel Dynamic Acceleration is being used, that can go up to 10.0 on a single core while the second core is in the C3/C6 sleep state. During normal use, the multiplier will constantly bounce back and forth between 9.5 and 10.0 depending on whether 1 or 2 cores are active and being used.

At full load, this CPU should be able to maintain the 9.5 multiplier regardless of what you are running on it. Anything less than 9.5 when at full load is throttling.

The easiest way to increase battery life is to reduce your core voltage. If you want maximum power savings at idle then you can try the Power Saver feature in ThrottleStop but I've found that it doesn't make that much difference in actual power consumption at idle. Most Intel 45nm CPUs go into a low power state at idle no matter how you have them setup so a super low multiplier and VID doesn't seem to make much if any difference at all.

You can always run the Windows Performance Monitor program and set it to monitor your battery usage. It will tell you what settings work best for your computer and what settings don't make much difference at all. Lowering your screen brightness a notch usually does more to conserve power than a lot of the user controlled CPU power saving features do at idle.

The Dell throttling schemes will make your laptop feel sluggish and unusable. At first they save power at the expense of performance but too much throttling can result in wasted power because your CPU has to spend more time awake spinning its wheels. If it was allowed to run at full speed, it could get the job done quickly and then re-enter one of the C3/C6 sleep states where the core voltage drops down significantly and you end up saving far more power. Throttling a CPU to 20% of its rated performance is not as efficient as letting it run at 100% of its rated performance for 20% of the time to get the job done and then letting it go into a low power deep sleep state for the other 80% of the time.

The VID info in the VID/FID request register is missing from the Core i design so simple under volting is no longer possible. Intel might have moved this info somewhere else but no one has discovered where it's hiding yet if it is still possible.

 

Here's another app (free) that seem to do what RMC and ThrottleStop do, it lowers the cpu frequency and the VID when needed. But its 100% all automatic.

http://grano.la/software/

 

Granola has 3 power profiles; Lowest Speed, Miser Ware and Highest Power. On Lowest Speed it is locked at 800 MHz and the lowest VID at full load or at idle. At highest power it is locked at full speed and the highest VID, even at idle. You can already do that in Windows so there is no advantage to using this mode.

There is no VID voltage adjustment in this program so the only option to try was MiserWare. This option will significantly slow down your computer but does that really save any power if your computer needs to be on longer to get the job done?

Here's a comparison to ThrottleStop.

Test 1 used Granola in MiserWare mode. To run a single thread of wPrime 32M, it took 112.8 seconds and was consuming approximately 24.2 watts while benching.



By using ThrottleStop instead, power consumption went up to approximately 25.5 watts but performance was significantly improved. The wPrime 32M bench was done in 82.1 seconds.



At idle this same computer consumes approximately 16.0 watts DC.

Granola:
24.2 W x 112.8 s = 2730 Ws

ThrottleStop:
25.5 W x 82.1 s = 2093.5 Ws

Slowing your computer down makes it consume significantly more energy because you will have to have it on for a longer period of time to get tasks done. Running a computer fast and then turning it off or getting it back into a low power idle state as soon as possible saves energy. By running a program that deliberately slows your computer down, you will only waste your time while your computer consumes more energy.

 

Thank you for that unclewebb, running my P8700 at multiplier x10 and 1.0250V right now.

To make throttlestop autostart i put it in the startup folder, there's no option in throttlestop for that right?

 

The startup folder or the Task Scheduler works for me. I didn't add an option to ThrottleStop for this because I don't like writing anything to the registry so programs are easier to move or delete if you don't want them anymore. For anyone using UAC, the Task Scheduler is the only method that works.

 

Quick question guys

Can I use RMClock to undervolt the Intel Pentium M, in my dad's laptop?
It's currently running XP


Thanks guys

 

Yes, you should be able to.
All the processor models that support Dynamic P-state transitions can be undervolted:
RMClock Utility. Products. CPU Rightmark

 

I've been trying to undervolt my AMD TL-58 on and off since summer of 2008 and I've been able to go down to 0.975 V from 1.075 V and running stable with Prime95 for hours.

The weird thing is that I can get BSODs and random reboots (with no BSOD or any other message) while the CPU isn't stressed at all. So I can be playing Oblivion for a few hours without any BSODs and then when I'm browsing the net or doing anything else which doesn't stress the CPU or any other component very much, I can get a BSOD or a reboot without any message at all.

Anyone else having this issue?

PS
I have tried all possible voltages between 1.075 and 0.975 with the same results.
I have tried undervolting on both Vista 32-bit and now Win 7 32-bit.

 

Thanks alot, I will give it a go

 

Great guide, though I have a problem:

The RMclock program that I downloaded does not appear to have the same options as the one in the screenshots.
There are no "Advanced CPU settings".

I'll post some pics of the issue:





Also, here's how hot my CPU got:

 

If "Advanced CPU settings" does not show up in RM Clock, that's usually a sign that your CPU is not fully supported. It has not been updated in almost two and a half years so it doesn't fully support the newer CPUs.

 

Do i need to make any changes to work with Windows 7 64bit system.

 

For Turion X2 we can use K10stat to undervolt

 

Guys, I have read most of your replies to this thread and was wndering about a few things that you leanered people may be able to answer for me...

The reason for my quest was that I have a Aopen notebook 1555a with a Phoenix BIOS, Pentium M 1.4Ghz socket 479 processor.

For the last few years I have been seeking a way of moving it from being sey at a 6.0 multiplier to the 14.0 that it should be capable of running at.

I have no way to set anything in the BIOS to turn on or off Speedstep or to change any settings regarding power etc there.

Subsequent upgrades from XP to Vista and now Win 7 have not helped me either.

I have played endlessly with power settings to try to get the CPU to change with any load and it still sits at half throttle, regardless of anything I can do..

I have tried numerous hacks like trying to run speedstop management toools from Intel for Win 2000 in compatibility mode, loading Hotfix CPU patches from various sources regarding early Centrino issues, Intel CPUID software says the CPU has been tried to be hacked and will not load...

I have run CrystalCPUID with no resltant changes and now tried the undervolting methods and reverse with RMClock to see if I can get a result and still come up without a sausage...

In a nutshell, I am a little frustrated...

Is what I am experiencing common with the older Pentium Ms and Speedstep Rnot responding or is it that I have a dud CPU?

PS:Tried to reset the CPU with RMClock and on the profiles area, the "Use P States Transitions" areas are all greyed out... Is this a sign that the CPU is not supported by RMClock does not support any other settings of the CPU state other than software?

Please help!

David Craker

 

I have installed rmclock with Windows 7 64bit. I am getting error related to "Administrator" and/or "64bit load error" while starting the application. Do i need to copy 64bit driver uploaded in OP post for windows 64bit as well?

 

Copy the signed driver from the first post into your RMClock directory and that should fix the problem you are having.

 

Thanks for the info. Now, It is working fine but showing wrong values. RMClock is showing CPU load above 50% when it is actually less than 10% according to task manager. It is showing core temperature around 35c when it is around 50c according to HWMonitor.

 

What CPU model do you have? Paste a CPU-Z image if you can.

By default, HWMonitor uses the wrong TJMax value for my T8100. For most of the laptop CPUs you can look up the correct value on the Intel website and then enter that value into the hwmonitorw.ini file. I think you can also edit a registry item for RM Clock so it uses the correct value as well but I haven't checked this yet. Without knowing this value, software can't determine the correct core temperature and for a lot of CPUs, Intel didn't do a great job documenting the correct value so software was left guessing.

 

Uploaded CPUZ screen shot.
http://i814.photobucket.com/albums/zz63/ritholtz/dv6-1353cl/cpuz.jpg

 

Intel lists the TJuntion as 90C for a T6600. HWMonitor likely uses the default value of 100.

You can edit hwmonitorw.ini and adjust this value:

CPU_0_TJMAX=90.0

 

WOOOOHOOOO

Finally cracked a solution...

Firstly, I formatted everything and went back to XP, then installed the XPSppedstep tool and that did not work...
I changed the power profiles to max but that did not work...

I was almost to bits...

Decided to do a wide search using Google of hundreds of weblinks using Aopen 1555 bios updates and low and behold, I found a Beta BIOS Update (Unnoficial) ona Russian website for my Aopen Notebook with a newer date than the last and latest official one of the Aopen websites...

I flashed the BIOS with a USB Boot key and the update and amazingly the processor is back to the right speed straight way.....

Now I have reinstalled Win7 and it remains at the correct speed now...
I have just ordered a new upgrade for the CPU to celebrate...

Cheers...

 

How do i fix issues in rmclock. According to task manager, my cpu load is under 10%. But RMclock is howing it above 50%.