The "Undervolting" Guide

The amdk8 is the default driver for your processor. That's the one to disable. There was never a reason to install an older driver, with all the bad consequences, when you already had the current one. The driver provides the interface between the hard-coded processor and the fluid operation of the OS. You can't modify the CPU, but you can change its comunication with the operating system. That's what you want to do.

 

I don't think it's the chip. I think it's because of your OS. If I recall, you have W7. When RMC was finished (ver. 2.35), Windows 7 didn't even exist. How could the developers have forseen any problems? I have a 45nm T8100 and it runs at 0 to 1% load on XP.

 

I have a T6500 CPU. It has a Power range of 1.00 - 1.250V. Does this mean I shouldn't run it below 1.00V on my highest multiplier?

 

There are no rules. You run your machine with your processor and your operating system at the lowest reliable voltage for the clock speed you want to run at. Sorry, no magic here.

 

no i have XP. The refresh rate of the graphs eats up WAY too many CPU cycles.

 

OK. My mistake. I turned my sample rate to 10 seconds (2710h) and the temp. graph shows a neat sawtooth sine wave of the fan turning on and off.

 

Ok, I did the undervolting on my MS-1651 with P8400. I have the lowest VID (0.875V) for all the states. It's rock stable. Now, since 0.875V is the lowest I can go, should I leave it as is? I mean I've run ORTHOS along with Furmark for 3 hours straight and it doesn't have a single problem. Also, MSI has these ECO modes that you can use. Let's say I use Turbo Battery, will it override the settings I have for RMClock? I see that RMClock added an entry into the power option, dunno if I should use it.

 

You're almost done!

1.) The P8400 is a nice one to work with. You answered your own question: leave them all at 0.875 and be happy. There no more to do.

2.) You want only RMC calling the shots, power-wise. Set the power option to "RMClock".

3.) In RMClock, create a Custom power plan in the profiles you will be using. Voila, every power parameter is under one application: Voltage, clock speed, screen off, HDD off, standby, etc.

 

I found my stable/optimum voltages on my notebook a while ago on Vista, but since then I have wiped out my hard drive and installed Win 7.

Am I right to assume that I'll have to go through all of this again and that the Vista-safe voltages won't be the same on Windows 7? In other words, I'll have to re-optimise rather than just plug in the old values I was using on Vista.

 

Seems that Windows 7 is too new for RMClock, so not all the power saving option are available under RMClock. That was easy to fix, just change the individual settings under Power Option for the RMClock plan. I don't think RMclock can overide settings it doesn't know, right?

 

Not supposed to do it that way. What are the items in W7 that RMC doesn't cover and do you need them. Not good to have 2 cooks in the kitchen, from my experience. At least make the common parameters identical.

 

The thing is i dont really think the k6/k7/k8 are drivers in the common meaning of that word(k8 is the newest and the k6 is the oldest). I had all three of them in the /system/drivers folder before the installation of the amdk7, but it doesnt matter really, because i tried to do the same thing in the registry with amdk8(the only one there) and nothing happened - no BSODs, but Powernow! seems to be still enabled (highlighted in crystalcpuid) and undervolting still freezes my ntbk.

I assume that the amdk7.sys is responsible for powernow, not k6 or k8, at least thats what i read somewhere on the net.

 

Hello people, one question.

I have a notebook (ASUS W7Sg) with temperature issues and undervolting wouldn't hurt... there are two problems.

1. I don't trust the default Voltages that RMClock gives me. They don't match the voltage specifications on the CPU datasheet page:
http://ark.intel.com/Product.aspx?id=33917
(unless I'm reading those incorrectly; the voltage on minimum multiplier is lower than the minimum voltage given in the specs; and the voltage on the max multiplier is larger than the maximum voltage in the specs)

So, the question would be if anyone here knows the correct default voltages for a T9300, per multiplier (or at least for minimum and maximum multiplier; interpolating between those should give good results).

Similar values for a T9550 [ http://ark.intel.com/Product.aspx?id=37130 ] wouldn't hurt either, I would then be able to undervolt my F6Ve --- although it doesn't suffer too much from temperature issues.

2. I also don't really trust the behavior or RMClock on these newer chipsets/CPUs. I don't remember details now, but I saw some inconsistencies when I last tried to use it. I would use CrystalCPUID but that software doesn't give ANY default voltages at all; so with that I return to my original question about the default voltages.


PS: I am selecting "Mobile" CPU manually under RMClock, since it detects it as a desktop CPU, due to chipset incompatibilies I presume.

Thanks in advance.

 

You can also try CPUGenie, but that one will set you back $14.95 after 30 days of use.

What I do is use CPUGenie to run the overnight test (it has an automatic test to determine the best voltages), and then plug those values into RMClock.

 

Default voltages are just that: Defaults. Suggestions. Hints. There are no "correct" numbers. Undervolting requires that you experimentally determine the lowest reliable voltage for each clock speed. Do you really think Intel is going to do that for each and every CPU they send out the door?

RMClock always defaults to Desktop. When you change it to mobile it just reads from a different table of numbers.

 

I don't think I made myself understood correctly.

What I expect is that each type (or family) of CPUs has some rated, default voltages (per multiplier?) at which it is guaranteed to work properly. And these voltages are used by the laptop manufacturers. If such a specification does not exist it would surprise me very much. Are the manufacturers responsible for inferring correct voltages on their own, then? For instance, the voltages that RMClock read for my Pentium M on my M6BNe were quite different from those given for my Core Duo Merom on my V6J.

What I need is these defaults, as starting points. I am definitely not asking for any "correct" (undervolt) voltages. I will of course determine these on my own, by decreasing in small increments and testing.

Or in other words, what should be the starting points for undervolting? Shall I just believe what RMClock gives me? Like I said, they do not seem to match the specs given by Intel.

 

no.

In RMclock select "no management", then go to the monitoring page and see your voltage.

or use some other software to see your voltage.


EDIT: even better, close RMclock and use some different softare to see your default voltage.

 

Yes you did. Don't be an NBR Procrastinator and start testing.
Or, do you want to be one of those "spoon feeders"

Does it really matter what Intel and RMC suggest? It's your laptop.

60 seconds of Orthos will tell you if you should proceed to the next step or not.

 

Ok. I have everything tested and have my voltages. RMClock is set to run on startup and I've tested and it works fine. I get the UAC box on start-up and answer yes and all is good.

Now, how do I set this up to start automatically when my wife logs on to her user name? She is just a standard user. I have restarted the computer and RMClock does not start automatically when her name is clicked. The program needs to be started manually. Is there a way to do it where it automatically starts?

 

Thanks, that's a good idea (wonder why it didn't cross my mind?!). Anyway, RMClock and CPUZ 1.50 read rather different voltages... e.g., at 10x 1.188V RMClock versus 1.300V CPUZ. Weird. I downloaded CPUZ 1.52.2 though, and that one agrees with RMClock. I will take that as 2 votes out of 3, and start on the basis of those voltages.

I doubt whether 60 seconds are enough... on my V6J I went through several iterations of Orthos with no errors, only to have the computer bluescreen later... I upped the voltages .1V and I had no bluescreens since. Orthos is not infallible.

How did you set up RMClock to start automatically? It has 2 ways, via Startup folder in Start menu and via registry key; you should use the latter. If you used the latter then maybe you need to re-do the same steps from her account. Also check explicitly that RMClock reads the customized voltages from her account.

 

I have it set-up to start via registry key. I re-did all the steps and I checked everything in RMClock and it matches what I have set-up in my account.

Could that she isn't an administrator and just a standard user have something to do with this?

 

Probably, almost certainly RMClock needs admin privileges to do the things it does... either move her user to the Admin group; or otherwise try to make sure RMClock always starts as admin (not sure how that can be done).

 

Ok, I managed to find some stable voltages starting from those given by RMClock (which concur with those given by CPU-Z 1.52.2 and 1.41, not with 1.50 -- probably due to the bugs mentioned in the main post).

But now the problem is that I have a drop in performance when going from:
a) unmanaged (default speedstepping) to:
b) managed with RMClock

On the T9550, this drop is about 5%, on the T9300 larger, 10% -- quite significant.

I am not sure whether this is due to:
1) Lack of halfmultipliers in RMClock.
2) Improper handling of IDA in RMClock.

For 1): the speed read by CPUZ is the same whether managed or unmanaged, so unless CPUZ doesn't know about halfmultipliers either, this is a point against 1).
For 2): I have tried enabling IDA in CPU-Z, but I can't seem to get it running: whether I use the IDA tickbox or the maxmultiplier tickbox, the real-life processing power is the same. Weirder, when I force IDA the voltage is changed to the default, nonundervolted voltage for max multiplier -- not to the voltage for IDA. So something fishy is going on.

Any insight/experience with these issues? Thanks.

 

I had to move her to the Admin group. Once I did that I was able to set everything up and it now starts automatically. Thanks for your help!

 

Results from another M1710, this one with a T7600. 14x [email protected](stock-1.2000v). Temp went from 83C to 73C.

 

Did you test EVERY FID for two weeks? If not, start over, hoser

 

So, while there is no permanent damage, here are my experiences with BSODs from my limited time with undervolting:

1) It took forever to boot-up from the BSOD
2) On the user screen, the webcam didn't initialize until much later than usual (takes 4-6 seconds usually, took over 20 this time)
3) Turned my wireless radios off (just had to turn them back on, though)
4) Forgot the size of my icons on the desktop

But nothing else (or anything I've seen), lol.

Currently testing 1.025V on my T9600.

~Ibrahim~

 

I did test EVERY FID I USE for two hours. I do not understand the need of some to test for days. I feel it puts undo stress on the system. I have not had a BSOD on any of the units I have undervolted, once I had settled on a final voltage for the FID.

I'll leave the endurance testing to others, thanks.

So, when are you and the Plexiglas polish coming to the French Maid? I have a jug of Brasso stashed, just waiting.

 

Due to my short attention span, I never tested for more than a minute per FID
When I got my first BSOD, I raised it by 0.025V and left it there for the last 2 years.
More interesting were the speedswitching experiments which you participated in.

I'm having my 8" belt buckle fabricated at Orange County Choppers. But I have no tatoos. Will I blend?

 

Got my Dell Studio XPS 13 (1340) undervolted. thanks to this guide & others.
P9600 2.67GHz

Stock voltage is 1.1625 across the board (6x - 10x), idles @ 40, full load @ 80

6x - 0.925 (Minimum voltage available, disabled this as 7x is same vid)
7x - 0.925 - idles @ 38 degrees
8x - 0.935
9x - 1.000
10x - 1.0625 - full load @ 70 degrees

saving 10 degrees worth of lap burn

 

Hey. I noticed you have a T9600. Did you use the half-multiplier registry hack to be able to undervolt the half-step multiplier of this thing?

 

Yes, the test was interesting. I was a little surprised at the results.

I do not know about blending, but I do know you will bleed if we get you a tattoo, as a memento of the occasion.

When will the buckle be done? Did you give them your avatar so they would have a pic of the DeLoren to use as a pattern? Hope it turns out okay. Can't wait to see it.

 

Is this the .reg file that comes with RMC or something else?

 

Further digging on my part leads me to believe that Intel implemented speed switching and throttling as part of their statistical process control program. In manufacturing, where you can't rigorously test every single unit, you have to make allowances for natural variations in your product. Plus, they have no control over the design and construction of the computers that employ their chips. In the case of an overheating situation, of course Intel would get all the blame! So, they ship the chips with stock voltages higher than necessary to prevent BSODs and a built in a safety valve (speed switching and throttling) to prevent damage to their chips in someone else's machine.

Some very clever marketing people apparently couldn't resist claiming that it was an "enery saving - battery life - fan noise" green program all along. But the numbers don't add up. I'm all for low power chips, low power peripherals, and battery longevity, but give credit where credit is due. Lower the voltages on everything and crank the clock speed up so Moore's Law will continue to hold for a few more years.

As far as the power calculations, the results are just pure physics. The experimental results match the mathematical predictions.

 

No, actually, I didn't. I just used Scott's method of disabling IDA in the Advanced CPU Settings page and then just using 11x as my multiplier. 11x (which is in the IDA spot technically) becomes 10.5x because that's the maximum for the CPU. It works out great, except for the fact that I lost IDA: sometimes, it would spike to 3GHz.

I've been meaning to try the registry hack, but can't find the time. And that guy's instructions (from the post I have a few pages back) are pretty vague.

Quick question: so I'm using RMClock Power Management as my power plan. BUT, every time I restart, it goes back to "Dell Recommended". What the crap? Is this a known issue?

~Ibrahim~

 

That would be a Dell driver somewhere. The majors always seem to load their own junk on their units. I guess they don't think M$ piles on enough in Vista

I've spent 5 years scraping XP down to bare metal. Runs all my stuff, fast and lean.

 

I looked up the "hack" you mentioned. It was first provided on the Russian Rightmark forum by one of the remaining guys on the team. He points out that it's just a visual trick (like my method). RMC can handle half-multipliers, but it can't display fractional numbers on the tabs. So, you shuffle the tab numbers around so that selecting a 10X instructs the program to try for 11X and it lands on 10.5X.

"Hey, want a green shirt? Turn on the green light!"

 

can I use that as my sig?

 

Did you choose "mobile" in the "cpu type selection" ????

Cause my original voltages were 0,875V for the lowest FID and 1,05V for the highest...which is actually lower than your cpu undervolted...Not logical.

I'm asking that cause if i don't choose "mobile", i have 1,1625V just like you...

 

Only if you promise to use it

 

Undervolt this, you hoser

http://www.youtube.com/watch?v=J0R5DTHcmGU&feature=related

 

I applied the 1 minute graph refresh rate tweak for RMC, but I stil get 2-6% CPU usage of RMC. Why is that?!

 

Usually, refresh rates are done in seconds or miliseconds. So either 60 or 60000. Make sure it is the "refresh every x amount of time" and not "refresh x times per second/milisecond".

~Ibrahim~

 

Yeah, actually I already timed it. It is indeed 1 minute.

When I close my browsers, only the task manager is opened, I get 0% CPU usage. My problem now is, when I move my mouse pointer anywhere in my desktop, even just a small move, the CPU usage booms to about 10% with explorer.exe taking up that 10%.

What could be the problem with that? I'm on W7 x64 RTM

 

What's your clock speed/multiplier settings? If you run at a lower clock speed, lighter things with trigger a higher CPU load.

I assume Aero's on? What graphic card do you have?

~Ibrahim~

EDIT:

Final voltages for T9600! Stock was 1.125 (71C load). Now have 1.0125 (70C load). w00t?

 

Here:



and yes I have Aero turned on, so that's the problem?

I have a 9800M GTS vidcard which I think is more than enough for Aero to be used.

 

Try checking Index 2/3. It will raise the clock/voltage some, but it may lessen the load. Just an idea....

Yeah: definitely should be more than enough.

~Ibrahim~

 

Oh ok. But does Aero use CPU resources?

 

I think so.

dwm.exe (transparency, Flip3D, etc.) is always consuming CPU cycles from me.

 

applying this principle to my desktop till my m17x arrives, might not bother with it, 10minutes into ORTHOS and 58 C max, surprised it runs so low, will be another story on a laptop, thanks for the guide!