The "Undervolting" Guide

7.0x was stress only 45min, 8.0x and 9.0x was stressed for 2hrs each, 10.0x was stressed 3hrs.

I only got BSOD once when i dropped the 9.0x down to .8975v
My Orthos stopped and my motherboard made annoying beeps when I drop my 10.0x voltage to 0.9375

I do I have SuperLFM I was just wondering cause all my lower multiplers were at the lowest. It's ticked now.

 

I should stick my specs in my sig, they're in my profile but nobody thinks to look there.



There's my CPU info. Apparently that version of CPU-Z reads voltage inaccurately but the rest should be accurate. RMClock detected all those voltages and multipliers (with the exception of the ones I lowered and tested) when I configured it like the tutorial indicated so I'm inclined to believe it detected it properly, even though my SuperLFM and 6x multipliers are wonky. I'm running at what appears to be 100% stable (I won't be sold on it until I run a couple extended gaming sessions, just like I did with my video card OC) but for now it appears to be functioning just right.

 

Netkiller: I see, fair enough, the beeps are a fail-safe measure

Though you might wanna test your 7x abit longer or raise it a tiny bit higher, cause that multiplier is the most sensitive to voltage changes. Maybe an hour or two test...

Lets say when you do a virus/adware scan, it usually uses the 7x-8x multiplier at those loads and it usually takes an hour to finish. If these multipliers dont have enough voltage, it will just BSOD in the middle of scanning.

Jlbrightbill: Yeah that would be handy if specs were on sig.

You need CPU-z version 1.41 for the correct voltages. I dont know how the recent versions got buggy but rmclock is accurate.

Yeah test it in real time to confirm 100%, prolonged gaming is a good way of testing it. Dont blame RMclock straight away if you get BSOD. I got a couple of BSOD's before and it turned out to be the video drivers -.-" .

 

1V stable on the T7500 ? I guess my 1.13 is high then

 

Would this work on the G1S considering it down clocks itself? I realize there is a difference between down clocking and undervolting, but the G1S down clocks to save power. So wouldnt that mean it UVs as well?

 

No thats not undervolting. Thats just downclocking to save power, yes it also means lowered voltages but its not 'undervolting'. The lower the clockspeed the less voltage it uses by default. Each clockspeed/multiplier runs on a default voltage defined by the CPU.

Undervolting is lowering those default voltages to a optimum amount

Yes it will work and benefit on a G1S, just like the rest of us

 

The voltages I posted for the X9000 earlier are wrong - I notice they were taken in "desktop mode". For mobile mode I once again get 1.2000V for x14 and 1.0000V for x6.
Ill just try undervolting with the mobile ones, even if they are wrong. Not much can break right?

 

Think of it as a lightbulb.. not enough power, no light. Too much power.. kablamo!

best analogy i can come up with...

 

hmm yea I was surprised how high my temperatures went though when I pressed apply on the desktop voltages. Lucky nothing broke :-/
Usually using that stress test from your guide i get max 55-57C, with 1.3125V I got max temp 67C.
I then (still didnt know it was desktop voltages) undervolted and got the 1.3125V down to 1.2000V. It wouldnt go lower without errors

 

Hmm ... my Dell D420 w/ a ULV 1.2GHz Core Duo would lock up when I try to run any programs if I start RMclock automatically first when windows boots up and turn down the voltage to .935. If I kill RMclock it would unlock itself.
Yet if I run my programs first and then turn on RMclock with the undervolting after I use the computer for awhile everything is fine.
I wonder if it could be related to the on-die cache it uses to cache the programs? Or just RMclock messing with Windows?

 

falcon: its CPU's also have a high voltage tolerance before they burn out

b534202: well its already a ULV processor.. sounds like software conflict though

let it run on startup using the task scheduler, but let it run on a delayed start up setting so it doesnt startup as soon as you go to windows

 

Is it possible to undervolt GPUs as well?

 

No as far as i know, you can only underclock. Its not recommended you tweak around with the GPU's voltages

 

There are some possible disadvantages of this. As a power supply engineer I can help shed some light on these less obvious problems. let me try to put this in an understandable way. As you lower the output voltage of a power supply you change the way it handles changes in load demand. We call this a load transient. This is a drastic change in the load on the power supply. When you lower the output of the power supply you can lower the gain or phase of its internal error amplifier (the part that keeps the voltage right under all load conditions). This can cause the power supply to become unstable under these heavy transient load conditions. This can cause seriously damaging ringing or oscillation in the output of the power power supply. also, you will increase the loss in the power supply and lower its efficiency adding heat and stress to it, most often shortening its life. the problems you encounter when you lower the voltage too low could be related to the power supply becoming unstable. I'm not saying that undervolting / undervoltaging (more correct industry term) will be completely bad or damage anything. most power supplies are built robust enough to handle the extra stress and if designed correctly "should" be stable at any voltage it is capable of producing. This isn't always the case. I wouldn't have a job if all my customers could correctly design a DC-DC supply.

 

We are only lowering the CPU's input voltages not the whole notebook. A voltage of .200 less to the CPU isnt something to worry about.

How about the notebooks system fans? They only go on when needed. wouldnt that also decrease and increase the transient load of a power supply?

 

Besides, the notebook itself raises and lowers the voltage by default, all we're doing is lowering the defaults.

 

There are several different "power supplies" in your laptop. All doing different things. the CPU power supply is always a dedicated DC/DC usually running off the system 3.3v or 5v buss. As you use more and more current a small amount of loss turns into a lot. The CPU is the most demanding component in the laptop drawing many many amps.

 

Exactly my point. Intel Speedstep technology was made so it can dynamically lower its voltage and clockspeeds. It was designed to have fluctuating voltages. Our power supplys would have been long gone if that was true..

Thanks for the warning though... except it doesnt apply for notebooks as the CPU was made to handle different voltage loads

 

All CPUs are made to handle different voltage levels. They're not designed to handle fluctuating voltages. they're made to be able to run at varying voltage levels. these transients Im taking about are in current demand and how they effect the stability of a DC/DC at different voltage levels. Its about the DC/DC having to drop from 3.3 to .9 instead of say 1.2 v and how that effects the stability of the control loop. the lower the DC/DC has to convert the voltage from the original the more loss there is in components like Mosfets and Inductors. I don't have the time or the patience to explain analog DC/DC engineering. theres a lot more to it then you think. if you are an aspiring BSEE major feel free to read up on my companies web site. If you just got your BSEE, live in the bay area and need a job. Hit me up. Were always looking for engineers.

http://www.linear.com/designtools/app_notes.jsp

 

Admittedly I'm not an electrical engineer, but every notebook has fluctuating voltages right out of the factory. I'm not seeing the problem here.

 

Its not the fluctuating voltage. the running voltage of the system is quite stable. usually aimed at 1% peak to peak ripple of the desired voltage and in many cases 3% peak to peak under 25% transients.. what youre calling fluctuating voltage is the ability of the mother board to change the operating voltage of the CPU. Usually so that the specific motherboard can handle a variety of processors or run at varying voltages to change the dynamics of the system such as power demand. whats important here is how you change the stability of the DC/DC at lower voltages and how that stability effects the CPU under transient conditions. I cant exactly tell you the effect of the lower voltage on the CPU as I'm not a processor engineer. Ive designed a lot of processor power supplies and know the effects on the DC/DC as you lower its operating voltage.

 

I think ill skip the theory reading and stick with the proven practicality of undervolting

If undervolting had this underlying flaw all this time, we would have seen a large amount of troubled notebooks by now. So far i failed to find a single person that can directly blame Undervolting w/ credible proof. Trying to find credible info about UV'ing doing damage was hard enough.. all i read was hearsay.. he said, she said...

Yes there is a very very very very small cases of trouble with undervolting but this can be ruled out as a defective cpu or an idiot user

Alot of people have been undervolting for years with no problems and with todays idiot proofed notebooks its harder to do hardware damage

So its really nothing to worry about..

 

I seem to have a problem with my RMClock.
I set it to run at startup via registry key,I set it to run automatically,I even set the scheduled task to run.
So it starts,I see the process in the task manager but I can`t see it neither in the tray or the taskbar.
If I want to start it again to modify settings or view the temps,I have to close the one in the manager and restart.
What gives ?

 

Okay thanks to a registry tweak I was able to unlock all the VID options so now I can set as low as .7125 if I wanted to. Using this I set my SuperLFM voltage to .8500, since I figured if flipfire's T7500 goes that low, mine should too.

I don't notice a difference though... when I'm looking at the CPU info tab it shows me at 600 MHz and 6x so obviously the SuperLFM is active, but at that speed/multiplier it still shows me at .9000v. So in that regard it's not using the .8500v I configured SuperLFM to use. When I go to CPU-Z (1.41, supposed to work properly) it shows my CPU at 1197MHz and still at .9000v which leads me to believe perhaps RMClock is saying SuperLFM is active but it really isn't. Any thoughts on this?

 

eleron911: did you tick 'run minimized on system tray' ?

jlbrightbill : the tweak didnt do anything if the voltage is still the same. the cpu probably automatically locked it in for safety

SuperLFM isnt detected by CPU-z with any version. Version 1.41 only detects the correct voltages. I dont know why they screwed it up on the latest versions.

 

Yea I did. And it`s not in the tray. But it appears in the task manager. So I don`t know if the voltage changes are running or not...

 

No they are not, it means RMclock hasnt loaded properly

Something is stopping it from fully loading. Try re-installing or cleaning your registry

 

Done that,uninstalled 3 times,used CCleaner for it every time after uninstall, still same thing.
I`m gonna try to start it minimised via startup folder this time..

 

weird... im running the test... cpu-z shows my VId as being 1.225V when in RMclock I have it at 1.1

 

I trust RMClock since the temps have indeed decreased.A LOT

 

Whoooa!? what reg. tweak? I've been trying to unlock all the VID options for a weak! do you mind sharing your method with me


edit: I figured it out, unfortunately intel has me locked to .95v and above.

 

so...whats the registry tweak? i wanna set this as low as i can go.

 

I dont recommend you try and tweak it any lower through the registry, this could cause big problems

lemm4: You need CPU-z v1.41 to display the correct voltages. The latest version seems to be buggy

 

flipfire: Thanks...

Question... what's more accurate... everest or rmclock? there's a 5degree Celcius difference between the two... everest is 5degrees higher on average...

 

Id have to say RMclock since its a dedicated CPU utility

I suggest downloading other temp programs like HWMonitor and CoreTemp to see if its consistent. You dont have to install these programs which is very handy, you can just run them off a .exe

 

Thanks.

In RMclock.. I get 32Celc idle...
In HW... I get 37Celc idle...

What's more accurate?

 

37c idle sounds more accurate to be honest. Unless its freezing cold where you live

did you get coretemp

we did get a few cases of temp inconsistency before but that was with a T5500 processor and RMclock was the correct one. Not sure why your getting it on a T9500

 

Yup Coretemp reads about 38c.... then again... also reads my penryn t9300 as a wolfdale xeon 5190 lol...

 

Excellent guide. For some unimaginable reason my 1.60ghz Core Duo T2050 was set at 1.2625v for all multipliers 6 through 12, which is terrible. At that voltage the cores maxed out at 76C and 79C, which are now 67C and 69C @ a stable 1.000v. If only I could set my 800mhz voltage lower than the minimum of .95v...

 

go into your rmclock folder there's a registry file called "tweaks" or something, run it.

go-->Start-->run-->regedit -->from the menu bar select "find" search for "unlockvid" when you find it, open it, and change the value to 1. restart rmclock and voltages down to .715 will appear.

NOTE: if your voltages were detected correctly you won't be able to go lower than you were

 

OOOOOOOH SNAP!!! looks like i've got some more testing to do...it unlocked my voltages all the way down to 0.7125.

 

Just as a bit of a confirmed further testing, 4 hours of Team Fortress 2 later, my T7500 at 1.0000v is rock stable. Previous maximum gaming temperature was 79C, the hottest I got during all this was 63C. Winner.

 

anyone try the t9300 below 0.95v (tweaked)?

Is there a dif between SLFM 6x @ 0.95v vs normal 6x @ 0.95v?

 

My T7500 runs at 1.12. But if you said 1.0 V is stable, I`ll try that and give Orthos a chance.

 

It's definitely worth at least a shot. Flipfire and I have slightly different stock voltages on our T7500's though--his 11x starts at 1.2500, my 11x starts at 1.2000, so I've already got a .0500 head start. I'd like to try bumping mine down a little further just so I can know my practical minimum voltage, I know 1.0000v is great but knowing your limits is always good.

 

SLFM 6x will idle at 600mhz. The normal 6x will idle at 1.2ghz

 

My 11x starts at 1.22 .
It seems odd that the same CPU has different voltages in our systems.
But then again we all have different laptops.
I`m currently running it at 1.07 without any issues,my temps have lowered even more

 

Asus F8SN-C1 T9300

6x-11x =0.9250
12x = 0.9750

Using Everest CPU Monitoring 100% load 12x
original temp 62C
UV temp 54C

tested by running Folding@Home 24x7

 

hmm thanks for the guide, I actually trying this on my BenQ S41 since I want to optimize the battery life. Strangely though, I got impression if Merom processor (t5250, 1,5 ghz) is not suited for undervolting. The first normal test with no undervolt resulting 77c on HWMonitor, when I reduce the highest multiplier (9.0x) from 1.2500 V into 1.1500V, it actually increase the maximum temperature into 77c ;

I'm still in middle of tweaking the highest multiplier, more to report later.

EDIT: it appears the 2nd test is not valid because HWmonitor seemingly still detect the tempreature from previous test.

 

Ahh Thanks...

Oh and what's the the difference between CPU clock and CPU throttle? Sorry I'm asking alot I'm new at this...