The "Undervolting" Guide

yeah, I'v heard of this before. Some capacitors are "whinning" or something. It means your motherboard uses bad capacitors, which vibrate a lot.
I have read about someone who used silicon to stop it, but I don't think it's a good idea.

Have a search of this forum, many users have had the same problem.
I feel bad for you, hopefully someone will be able to help.
If not, then maybe you should start a new thread?

 

is it just me or does CPU Clock have a few problems holding onto settings in Windows 7? My settings seem to randomly get lost.

the computer in question is an ultra-low voltage Core 2 Duo. It knows what it is and everything, and the undervolt works (as evidenced by a BSOD on my way downvolting). but it seems like there might be some Win7 compatibility bugs.

that, or I'm doing something wrong. (not my first CPU Clock install)

 

here is something for you:
http://forum.notebookreview.com/showthread.php?p=5123555#post5123555

 

Everyone should remember that RMClock was written 3 years ago before half multipliers and at the very beginning of Vista. That's a lot of distance from today in the world of CPUs and operating systems. Don't expect your setup results to match Flipfire's tutorial eactly if you have newer stuff. And don't complain unless you've personally written useful software and have given it to the whole world for free.

 

It will help you find your BSOD point a lot faster. Then, raise up the voltage a notch and test more thorougly. No need to cook your CPU at the higher voltages where you know it's going to be stable anyway.

 

As you have found out, undervolting is for overheating not battery life. The power savings are small. Speed switching saves nothing. Real battery savings are from turning off the HDD and display when not in use. Standby saves a little bit more.

 

Before undervolting, the CPU idled between 39-42. Now it's 36-39. So it's running cooler whenever it's on, whatever I'm doing. Nice to see that it runs cooler across the entire range, because I don't stress the CPU to 100% very often. (I got the 9700 instead of a slower 25W CPU mainly to try and shave a little time off estimating statistical models).

 

Just to make sure, if your voltage for the max multiplier is stable, if you apply that voltage to all the lower multiplier, it should be stable as well? It makes logical sense but there may be some example to the contrary so I just want to make sure.

 

Measuring temps. at idle doesn't mean anything. The CPU doesn't consume meaningful power or generate much heat when it's doing nothing. Secondly, Intel has taken note of the overheating issue and has lowered the factory-set voltages on the newer 9000 series processors. They also wized up and reduced the number of FIDs down from 6 or 7 speeds like my T8100. If you're not overheating at max load then there's absolutely no reason for speed stepping or throttling. They're a waste of processor cycles. Somebody noticed.

 

Can you elaborate? Are you saying that if the CPU doesn't heat at max load at 1.0375v (at 10.0x), I might as well use that voltage for each FID? Is there a problem with the profile I created? Obviously, I'm new at this!

 

OK, I'll give you the short version. In Profiles, uncheck everything except 11X. Change the 11X voltage to say 1.05 just for starters. Select Max Perf. profile in all 4 boxes. Now, go to Max Perf. profile, and check 11X only (others will be greyed out). Now go to Adv. CPU and uncheck IDA and SLFM. That's it. Full speed all the time, no overheating, and no additional battery drain at idle. Later, you can try to see if you can lower the VID a bit. Maybe you can get as low as for your 10X. I did.

 

This is how my Advance CPU Settings screen looks like:



Should I adjust the highlighted parts to match the one shown in the instructions? BTW I have a T9300 if that matters.

 

Thank you so much for this guide! I spent pretty much all day undervolting my laptop and now I get 15 minutes more battery life on high performance and almost 35 on power saving mode. The CPU also runs about 10C cooler. Thanks again.

 

way to not in any way address my question at all.

I'll repeat it, for the attention-deficit: is anyone having issues saving profiles after reboot in Windows 7?

I'm gonna try running it in Vista compatibility mode to see if that stabilizes it. If it does, I'll let you guys know. it's really odd in that sometimes it does and sometimes it doesn't.

 

Actually, the grammatically correct term is attention-deficient. And for the reading-impaired, yes, lots of W7 users are having problems with RMClock

 

To this with issues with getting RMClock to work in W7, I believe that it should be pretty easy to get undervolting to work with Crystal CPUID and the task scheduler.

 

Running at lower frequencies and lower voltages at idle does save power. The reason for Intel to implement undervolting and switching to lower multipliers is to increase battery life, not to keep the CPU from overheating. And it doesn't take a rocket scientist to realize that if it is running hotter, it is using more battery power.

 

Intel didn't switch to lower multipliers, they switched to less of them. Their newer CPUs go faster than ever. Too many gears in the transmission wastes processing time.

When Intel lowers the starting voltages, it's called factory-volting not under-volting. After all, it's their chip and their call. They realized they were being too conservative.

"And it doesn't take a rocket scientist to realize that if it is running hotter, it is using more battery power"

Well, I am a scientist and I don't get this part at all. At full load, Orthos style, my CPU (at stock voltage) consumes 35 watts out of the total 100+ watts used by my rig. After undervolting, the CPU uses 25 watts. At zero load, (idle) my CPU uses about 3 to 5 watts. The display, HDD, chipset, memory and LAN use over 2/3rds of the power when running.

Running a laptop at a lower frequency does not prolong battery life because you'll have to run it longer to finish every task.

If you read some of Intel's published papers, you'll see that over-heating has clearly been a concern for them.

 

By switching to lower multipliers I'm talking about what the chips are doing.

Again, I'm talking about what the processors themselves do. They, thanks to software, can lower their multipliers and reduce the voltage to conserve power.

When my laptop is idling at 12 watts, which it spends most of the time doing, every little bit saved makes a big difference. Of course if you are chugging through something, it is optimal to have it at its highest frequency as at any frequencies these processors are ever going to see in a laptop, performance per watt increases with frequency and there is less time that the other hardware is sucking power.

Switching to higher frequencies and voltages when required and lower frequencies and voltages when not required does save power. How often are most people really using the CPU 100%?

 

No problems here with RMClock and W7

 

okay, I've figured out how to get it working smoothly with Win 7 (RC).

If you're having issues with it saving settings after reboots, turn off UAC (User Accounts Control). Disable it. For those that don't know, this is the function that makes Vista/7 constantly ask if programs have your permission to open. Something about UAC appears to affect RMClock's ability to remember what the hell it's supposed to do during a bootup.

Disable UAC and profit.

 

Intel has found you want to use the CPU 100% when doing anything.

It's far better to get a task done as early as possible and put the core to idle rather than work at a lower frequency.

But anyway I am being picky.

 

No you're not. You're being correct. Microprocessors were single speed for over 20 years until notebooks and their stringent cooling and battery requirements came along. A few years ago I had a !00 watt Pentium IV laptop that never, ever overheated. It weighed a ton, had a small 800X600 display and a noisy heat pipe cooling system that must have cost a lot to fabricate.

Now, it's all about light weight, big bright displays, high rpm HDDs, and cooling fans only a bat could hear. Packing all this stuff into ever shrinking cases is bound to cause heat dissapation problems. And, it has. Intel developed speed stepping and throttling as automatic protection systems. They had to set their stock CPU voltages conservatively high because an off-the-shelf BSOD on a new notebook model means warranty returns, recalls, lawsuits, etc.

I tend to agree with the folks who say that speed stepping doesn't save any energy because computers work in batch mode. A task run at half speed takes twice as long to complete. Energy-wise, it's a draw. I tested this myself with Super Pi. Yup, got the predicted result. By far, the biggest energy hogs are the display and HDD, at least at idle. I turn mine off after just a few minutes of non-use. I already know my "2 hr." battery will go days on standby and resume in seconds. I read somewhere that a notebook was measured with a volt/amp meter and only used 3 watts on standby. Works for me.

 

First, thanks for the wonderful guide.

Second, a question. I own a Thinkpad X300 and want to lower the average temperature. With RMclock however I don't get the possibility to select anything lower than 0.900 volt, and it is already 0.900 volt.



What should I do to get lower?

 

soo...any news on half multipliers yet? with a free program?

 

Tommy: that 0.9v is a factory setting, you cant go lower than that

kazaam55555: Use this guide, it works perfectly for me: Click

 

The guide in the first post could really use some additional info about Vista. Below are some things I've found that will hopefully save other people some headaches.

Automatic startup of rmclock:
The task scheduler seems to be the best way to achieve automatic startup of rmclock without UAC interference. See this post:
http://forum.notebookreview.com/showthread.php?p=4934336&highlight=scheduler#post4934336

Vista power plans:
Vista's default power management settings conflict with rmclock since both try to dynamically adjust the CPU clock and voltage. As a consequence, you may see your voltage bouncing up and down even when your cpu is idling at the lowest multiplier, or you may see the clock increase even when rmclock is set to the Power Saving profile. To disable Vista's dynamic switching and allow rmclock to have control, you need to edit the advanced settings of the power plan(s) that you use, and set the min and max processor state values to be the same. Note that if you have "OS power management integration" enabled with custom settings, do NOT try to directly edit the settings of the "RMClock Power Management" plan under Windows Power Options. That rmclock power plan is merely a container for the "custom" settings that you enter in rmclock, and it is deleted when rmclock exits. See the rmclock.htm documentation file that is included with rmclock for more info.

CPU modes not available in rmclock:
On some machines there are CPU modes available via Vista's native power management that are not available via rmclock. For example on my machine with a T5800, if I shutdown rmclock and set Vista to the default Power Saver plan, it appears to run 1.0V at 800MHz. The lowest voltage rmclock gives me is 1.05V at 1.2 GHz. If you are looking for max battery life, it might be worth experimenting with Vista's native power management to see if it gives you options that rmclock doesn't.

Note that when my T5800 is running at 800MHz, CPU-Z reports it as running at 1.6GHz. Based on tests with SuperPI though, I can see that the performance corresponds to 800MHz. This sounds like SuperLFM, even though the T5800 supposedly doesn't support SuperLFM, and RMClock doesn't offer the SuperLFM option on this machine. Strange.

Edit: Revised above info to recommend editing of processor power states in Windows power plan.

 

That's a shame. Thanks.

 

If this stuff has so many advantages , why doesn't the fabricant do it ?!

 

Same reason Honda won't install a Greddy turbo in your new Civic. Liability.

From last week here:

"Intel developed speed stepping and throttling as automatic protection systems. They had to set their stock CPU voltages conservatively high because an off-the-shelf BSOD on a new notebook model means warranty returns, recalls, lawsuits, etc."

 

i need some help with this. i did this a few months ago and had no problems. however, recently i got a huge problem. my settings on rm clock reset and now i cant undervolt it lower than 1.15 volts how do i fix this? (i reinstalled windows 7, formatted my hdd, reinstalled rm clock several times, reloaded the 64x drivers...im on 64 bit)

 

Re: Speedstepping and RMClock, I've found it effective to use maximum performance when plugged in, and power saver (using OS settings) when on battery. The machine idles slightly warmer - about 43-46 degrees as opposed to 36 to 39 when using performance on demand. That still doesn't seem dangerously warm to me (am I wrong?). And during heavy use, there's really no difference at all (no surprise there). On the other hand, I do notice a difference in basic responsiveness when the CPU is running at 2.8 all the time. It's nice

Performance on demand will eke out a little bit more battery life, but Vista power saver beats it. I can't get voltage as low using RMClock as I can by just letting Vista do it. The price is that the P9700 gets downclocked by 2/3. It's drastic, and it wouldn't make a lot of sense, except that almost 100% of my CPU intensive work is done at a desk, plugged in. On battery, I'm usually just reading / editing docs, checking email if possible, etc. Stuff that doesn't keep the CPU at load even when downclocked.

 

Did you choose "moblie" in advanced CPU features/intel core 2 family advanced features?

 

mhm i did.

 

alright, my computer bluescreened and for some reason (probably heat, i have been running massive video conversions for 3 hours and my temps were at 89 C when my antivirus decided to start scanning ) i can drop my voltages down to 1.0500 volts. which isnt quite what it was at. (i had it at 1.0125 volts) but its much better. im now getting max temps at 77 C so im pretty happy.

 

alrighty, another restart after downvolting and im down to what i had before.
my voltages are at 1.0125 and i get a max temp of 74C on the CPU, 81C when i give my gpu a massive overclock (500/600 to 615/756)

im happy again

 

interesting, to say the least

any idea why/ how it happened?

 

i have no idea. lol. im just happy it did happen. my best guess is dropping the voltages and restarting reset the minimum voltages somehow

 

Can anyone tell me if I'm doing this right?

Sony Vaio FW series
P8700 @ 2.53 ghz

screenshot
http://img29.imageshack.us/img29/4231/tempw.png

I don't know why but the voltage won't go lower than 9.25v

 

looks right.
But I don't think you need to run ORTHOS for 2 hours.

 

Yeah, everything seems ok, and also, long stability tests never hurt. The .925 v is a factory setting of your CPU, you cant go lower than that (as far as I know)

 

Has anyone experienced high CPU usage from RMClock when management is enabled? When I say high I mean enough to make your CPU step up to the next multiplier. If I compare the monitoring graphs when management is enabled and disabled it is obvious to see that RMClock is causing the CPU to step up to my top multiplier every second or so. If I turn management off it quite happily sits on the lowest multiplier. Kind of defeats the object of running RMClock.

I've also tried CystalCPUID and the multiplier management in that is not good at all. For no apparent reason it will just stop managing the voltages. If I then go back into the management screen and click apply is jumps back into action again. Maybe something to do with Windows 7.

Either way I can't use either program currently.

 

One of the more difficult concepts to grasp about about computers is that clock speed (FID or multiplier) and load are two completely different things. A multiplier is the clock frequency that your CPU "marches" to like a drumbeat. The load is the amount of electrical current racing around the transistor gates to make them process information.

My favorite analogy is your car. You can run the RPM way up while you're parked and generate very little horsepower. Why?. No load. You're not moving anything. RMClock does two very separate things. First, it allows you to run your CPU at lower than Intel-spec voltages at all CPU clock speeds. That reduces heat production, espescially at high loads (like a car going up a hill). Secondly, Intel intrduced speed-stepping to slow down the CPU under overheating conditions like down-shifting an overheating car pulling a trailer going up a hill. Speed stepping is kind of like an automatic transmission. It will automatically step up to a higher frequency as long as there's no heat problem.

Personally, since I've undervolted and can't break 67C at 100% load (gas pedal) at my highest multipler (RPM), I don't use it. First of all, I didn't drop a grand on my notebook to have it run at half-speed all the time when I need it to run applications. Secondly, as I've posted before, speed-stepping does not increase battery life. An application run at half speed takes twice as long. It's a push, energy-wise. The best way to conserve battery is to turn off the display and hard disk when you're not using your notebook. Standby is the best battery saver I've seen yet.

 

Thanks for the analogy Scott. It is however a little more complex than you describe. Firstly my lowest multiplier (x6.5) runs at a voltage of 0.925V. My next multiplier which is also my highest (x7.5) runs at 1.138V. I can undervolt this to 1.0V using RMClock.

As an interesting sidenote if I select a low voltage with an (X) i.e. unavailable in CrystalCPUID I can actually get this x7.5 multiplier to run at 0.925V. Just a shame CrystalCPUID doesn't always kick in to action.

Now going back to RMClock if management is disabled the CPU is running at x6.5 and 0.925V for the majority of the time. If I enable management the CPU is running at x7.5 and 1.0V for the majority of the time. That to me shows that it's pointless me running RMClock.

CrystalCPUID on the other hand does not have this problem - the CPU will quite happily sit at x6.5 for the majority of the time. It's just this other bug which prevents me from using it.

 

Hi there, thanks for the awesome guide. My CPU is a P7450 and RMClock did not entirely recognize it, however I don't believe it is an issue since it does not use half multipliers. It actually only uses 3; 6, 7, and 8.

Well it reports my 6x multiplier at a minimum voltage of 1.0 V. I managed to undervolt my 8x multiplier to the same 1.0 V. I guess I can't be any happier, right? Unless there is a way to lower the voltage more for the lower multipliers? From what I've read apparently thats not possible.

So since 8x and 6x are running at the same voltage, I should just run it full blast all the time?

Thanks,
Cam

 

If at half speed it was only using half the power this might be true but that's not the case.

Simplifying things a bit, if for example we run a task which takes 2 hours to complete using 6x multiplier @ 1.000V and the cpu uses say 10W of power at this setting then the cpu consumes 2hrs x 10W = 20Wh of energy. Now if we double the frequency using 12x multiplier @ 1.400V the task now takes 1 hour to complete but the cpu uses 1hr x 39.2W = 39.2Wh to do it. Even though it takes half the time to complete the task the cpu requires almost four times as much power or nearly twice as much energy to do it.

If we expand that a bit further to include 10W for the rest of the hardware and a 45Wh capacity battery then at half speed the battery would last ~2.5 hours. With the 12x multiplier the battery would last less than one hour and not even be able to finish the task.

Of course if the computer is idle most of the time then these effects will be far less pronounced and perhaps insignificant.

 

Actually this part is easy once you understand that the clock frequency is not the real gas pedal of your rig. Many posters have found Vista and W7 operating systems more "possessive" about controlling frequencies and voltages and often "steal" control away from RMClock if it's not configured corectly. CPUID reverts even more often.

Now to the point. You say you can get 6.5X to run at 0.925V and 7.5X at 1.000V. OK, the rule is the higher the clock speed the higher the voltage. And you say you dropped your top voltage from 1.138 to 1.000 which is a 12.1% undervolt. That's just about the usual amount people get here. I assume your CPU now won't overheat even at 100% load. Everything you said is predictable except one thing: Why on earth do you want to run your computer at 6.5X instead of 7.5X?

 

Thanks for the tip! My voltage was bouncing around a lot on Windows 7 and I couldn't seem to keep it locked... I didn't mess with any of this EXCEPT I made the min and max processor state match at 100% on my Windows power profile.. no more voltage jumping!

(not entirely sure what this feature is for, actually, as speedstep seems to work exactly as it did before)

 

When I do that my voltages and speeds are locked on the top multiplier (x7.5) and SpeedStep doesn't work at all.

And why would I want to run it a x6.5? To save battery when I'm not on AC as x6.5 runs at 0.925V.

0.0 brilliant post and couldn't be clearer that SpeedStep is very worthwhile when running on battery. However am I right in thinking that when on AC SpeedStep is pointless if you are undervolting i.e. you might as well run at your top multiplier?

 

That's what I do. I did notice that my notebook idles a bit warmer -- but not much. Anyway, if it's on AC, that means it's on my desk. And if it's on my desk, it's probably not going to be idle for very long.