The "Undervolting" Guide

I lied, I restarted and forgot to select Performance on Demand.

@ 2.2 GHz before I started my CPU got to 76C load with Prime95

Now with lowering the voltages from 1.3625 to 1.1125 (lowest voltage I can go before it blue screens) full load is 62C full load. And that's without my Antec 200mm big boy cooling pad (I'm at work right now).

 

I have an Acer Aspire 6920G with an Intel Core2Duo T7500 and undervolted at 11x @ 1.0000V. I know that some people could undervolt the same CPU under 0.9750V...

 

It all depends on the processor and if you get lucky or not. I get a blue screen almost immediately after Prime95 runs under 1.1125v. Besides I got the processor for free so I'm not complaining.

 

The T7500 supports voltages lower than 1.000 volt when it is using SLFM, Super Low Frequency Mode. RM Clock refers to this as Dynamic Front Side Bus Frequency Switching. If RMC has this option, make sure it is checked.



I've seen some T7500 CPUs go as low as 0.85 volts when this is enabled.

 

thanks unclewebb! had the same issue as Dimmuklok (stuck at 1v minimum even though RMClock showed that the minimum voltage could go lower, say 0.875v for 6x on my P8600) and just by ticking the DFFS option the issue was fixed!

 

Hey, I dunno if this has been asked an answered... i searched a bit n couldnt find any answer. Please accept my apology if this has been discussed, there r 500+ pages in this thread, and I admit I havent read them all.

I installed RMclock on a hp dm3, win7 64. added the signed driver (wrote over the one that was in the install directory). I can start RM clock, but the advanced cpu options is not listed. I cant follow the steps in the guide cause the tabs missing.

any ideas?

my hardware-
hp dm3z with amd k625, ddr3

 

If the advanced CPU options are not listed then your CPU is not supported.
RM Clock has been out of development for over two and a half years.

 

kk thanks, i thought it was weird that the cpu showed up as AMD engineering sample..

 

So I suppose any processors made in the last 2 years are not supported?

 

RM Clock partially supports some of the newer 45nm Core 2 CPUs but it doesn't fully support their half multipliers. It also doesn't support any of the newer Intel Core i CPUs like the Core i5-520M in your Sony Vaio. I'm not sure about AMD but any recent CPUs won't be fully supported by RM Clock.

 

Yeah I looked it up. Intel completely rewrote how that stuff is handled.

 

Well, this thread is only 542 pages long so forgive me for not finding my answer. Real simple question.

Does Win7 64-bit require the special drivers like Vista 64 does? The Original Post doesn't reference Win7 at all as I assume this guide was out before Win7 was.

I tried using the Vista 64 file and it seemed to install but then I got the error screen where Windows didn't know if it had installed all the wya or not.

I copied in the sys file and it seems to we working.... Just wanted to make sure I didn't miss anything,

 

Yeah use the same file as Vista 64 (the registered driver).

 

I wiped my machine and put Win7 on and couldn't figure out why my battery was dying so fast. Then I remembered I had to redo the undervolting.

I have a Gateway M6862 with a T5750 2Ghz chip (according to HWMonitor).
The stock setting was 1.250.

I was able to get the top multiplier (12x) down to 1.000 and the 10 minute test still worked fine. Any lower than that and it's fail right away. I bumped it up 2 notches to 1.0250 and we'll see how that goes.
Heat dropped from 83 degrees down to around 70.

Now the part that threw me is that when I unchecked the 12x and let it run at 11x, I was able to set that at the lowest setting available to me in Rightmark which is 0.950. It cranked through the 10-15 minute test with no issues at all.
Heat dropped from 70 at 12x to 61 at 11x which seems like an awfully steep drop.

I guess that means I can probably set the 10x - 6x ones also to .950 and see how it goes.

Would that be normal for just the top one to need a bit higher number and everything from 6x to 11x could run at the lowest setting? Great if it works that way as the heat should go way down and I guess time will tell.

 

I've been using RMclock for quite a while now but i never understood why my throttle clockspeed is ALWAYS lower than the cpu clock speed.

When i leave it on no management, than core clock speed matches throttle clock speed. But whenever i use performance on demand i can see that throttle willl perform lower than the normal speed why is this?

I have a t9600 C2D cpu with 10.5 * 266 mhz max = 2793 throttle speed on performance on demand max 2660 (something to do with 10 * 266 ??)

 

kelemvor: What you found is typical. To use the highest multiplier can sometimes take a significant increase in core voltage for it to be stable. Many users find that when on battery power it is not worth it and disable the last multiplier.

kaizah: RM Clock does not fully support the 45nm CPUs because of their half multipliers. Some of the RMC graphs like the throttle graph become inaccurate because of this on the newer Core 2 CPUs like your T9600.

 

I've been using RMClock for ages as well, but lately noticed that it really slows my machine down (HP8510P) compared to when it's not running. So I've decided to run without it for a while.

Now I'm using CoreTemp to check temps while running. With normal use it sits around 40-45C. Browsing heavy pages takes it to 55-60C. Haven't seen it go above 65C yet.

So is that temperature range ok for this laptop, in terms of longevity? With RMClock it was ~10C lower, which made me happy, but it just slows things down too much. Am I ok without it?

I guess an obvious question is - does everyone else experience a significant slowdown when undervolting with RMClock, or is this an issue with my setup?

 

Your temps are fine even without RMClock.

 

Thanks sgogeta4. And is a loss in speed the trade-off of undervolting? I was surprised how much Performance on Demand mode slowed things down. Strangely, not noticeable during gaming, but definitely things like web browsing, opening docs and .NET compiling.

 

antic: Are you using Windows 7 and what power profile are you using? There can be situations where RM Clock has to battle against the Windows 7 power management features which only wastes CPU cycles. Try using the high performance profile so RM Clock can manage things without any interference and that might work better. Your temperatures are fine if you're not interested in under volting.

 

Thanks unclewebb, I see.. it would be excellent to sort that out, if it's the case. I'm using XP SP3, processor is dual core T8100 2.1GHz. I'm always running plugged in to AC.

RMClock (v2.35.0 unregistered) settings which may be relevant:
"Use OS-Based Load Management" is on.
"Run HLT when OS is idle" is off.
P-state found at startup.
"Enable OS power management integration" is on.
"Use P-state transitions" is on. All indexes ticked.
"Use throttling (ODCM)" is off.
"Target CPU usage level" is 50%
"Up transition interval" is 200, as is "down transition interval".
CPU type is set to "mobile".

Does that look correct?

 

I haven't used XP for quite a while. All your RM Clock settings seem OK. Have you looked in the Control Panel -> Power Options to see how that is setup? How you have Windows setup might be interfering with RM Clock.

Have you run any benchmarks that show your computer is running slower or does it just feel slower? RM Clock does not fully support the half multipliers that your CPU uses but it shouldn't be slowing it down that much. Does the Task Manager show RM Clock using up a lot of CPU cycles or is any other program doing this? At idle on my T8100 with Windows 7, RMC shows 0% CPU Usage when it is minimized to the system tray.

 

Thanks again unclewebb. RMClock shows anything from 0% to 3% in task manager, minimised to the tray. I did a simple test where I continually refresh a web page in Firefox and watch CPU usage, while RMClock is active and then when I exit it. This is the result:

When I exit RMClock, CPU usage drops, and Firefox speeds up, refreshing the web page about twice as quickly. Something definitely weird there.

In XP Power Options, I'm not sure what could be interfering.. I've just got "turn off monitor" after 10mins, hard disk after 30 mins, standby is never (I'm always on AC power). Advanced tab just has "ask me what to do", except for "close lid" which is "do nothing". Hibernate is off.

Anything else I should look at?

 

I've just tried a couple of things, sadly to no avail. After reading that XP changes CPU management behaviour based on the "power schemes" dropdown, I tried it on "Home/office Desk" and "Minimal Power Management", but when I run RMClock it still slows everything down.

Also tried the "NoRealtimePriority" tweak for RMClock, which didn't seem to have any effect either.

 

With windows XP you must have "ALWAYS ON" on power scheme options.
Avoid using rmclock, it's obsolete software, no longer developed.
And if we look at your cpu usage, probably rmclock was using a lower multiplier (because you selected that option). Probably cpu usage isn't higher... your multiplier is lower... I would check CPU-Z to see the cpu clock speed.

 

How do my voltages look?

Core 2 Duo T9300

6x - 0.9500
7x - 0.9500
8x - 0.9625
9x - 0.9625
10x - 0.9750
11x - 0.9750
12x - 0.9875

Did this awhile ago but just getting around to posting this now XD

Lately it hasnt seemed much lower temp than without undervolting, so idk whats wrong or if its not actually active anymore or what. lol

What else could you use?

 

Use throttlestop. Take a look at unclewebb's signature above on this page.
Rmclock only works fine with very old cpu's (pentium m/ first core2duo) with newer cpu's rmclock WILL REDUCE your system performance.

by the way, you have nice voltages. Getting that cpu stable with less than 1v is excelent!

 

Only by 0.1Ghz though, as said in this guide, so thats pretty unnoticable and wont matter.

Thanks lol

I've looked at Throttlestop before, but it doesnt change your voltages like RMClock does

 

Throttlestop can change your cpu voltages, thats how I undervolted the p8400(gone, sold) and the x9100 in my ASUS.

 

Im looking but i dont see how to... can anyone explain it? are there any guides on it like with RMClock?

I noticed that with RMClock on, for normal usage it was locking it at 6x multiplier, when i shut down RMClock, it goes up to like 12, on the same usage amount. And then CPU usage goes down. I wonder if its been slowing down my PC like this for awhile...

 

Here's the testing I did that showed me that finely tuning the VID voltage for each multiplier on a 45nm Core 2 CPU is pointless.

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

A 45nm Core 2 CPU ignores the VID request register at idle. Using all of the intermediate multipliers makes a CPU inefficient and should be avoided. It's better to have a fast transition to full speed, let the CPU get its work done and then get back into one of the deeper C3/C6 sleep states where the real power savings are found. A slow CPU is not an efficient CPU. Microsoft must have figured this out too because Windows 7 is designed to get a CPU up to full speed a lot faster than Vista or XP does.

That's why ThrottleStop is different. It only lets you adjust the maximum voltage because that is the only important one to adjust. If you have any ThrottleStop questions, head over to the ThrottleStop Guide in my sig and post there.

 

Hey guys, as previously stated RMClock only recognizes very old CPUs. I checked out ThrottleStop, only to find that it's designed to work with Intel processors. I have an AMD Turion X2 Ultra (x64). Is there a utility out there that I can use?

 

There's a program called K10Stat for the more recent AMD CPUs. I don't have any experience with it but you can try that.

K10stat

I'm not sure where the official home page is or if there is a newer version.

 

ihaa, undervolted intel c2d P7350 from 1.1375 to 1.000 V \o/
I only ran 10-minute cpu stability tests, but:
on 1.1375 the CPU would immediatelly heat up to around 88*C and then overheat (max temp 90*C according to Intel)
on 1.0000... Well, after 10 minutes of testing, It oscillated around 82*C (sometimes going to 83*C, sometimes dropping to 80*C)
simply awesome, no more throttling! \o/

EDIT: Now trying with 0.9375... Let's see if it's okay with .200V drop. Will do further testing tomorrow, yay. So far 77-78*C

 

That program worked great for me! Thanks for the recommendation!

 

Is a black screen then an automatic reboot (once, it was a purple screen) another symptom of an inaccurate UnderVolt or would it only Blue Screen ?

Thanks

 

I would say it's an indication that you lowered the voltage too far. Sometimes my computer would not display a bluescreen, but would automatically reboot.

Other times, it would simply hang. One thing that I would advise is to interrupt the automatic reboot by turning the computer off and doing a hard reboot. Occasionally, my BIOS wouldn't find a boot device upon bluescreen + automatic restart, but powering off and back on fixed this. Other times, Windows 7 (x64) would give me a Startup Error and would ask me if I wanted to utilize Startup Repair, or try to boot normally. If I chose the latter option, it would try to boot, reboot, and give me the same screen. Powering off and back on fixed this, as well.

 

OK thanks, more thorough testing tomorrow then.
Appreciated.

 

Are CULV SUxx00 processors not supported by RMClock? Works fine on my Vostro 1500 T7500.

Under the Performance on Demand the P-state transitions only have 1 box to check and that is the 6x multiplier.

 

The SU4100 and SU7300 both use a maximum 6.5 multiplier. RM Clock does not support half multipliers so it won't let you access this multiplier without a RMC registry hack.

http://ark.intel.com/Product.aspx?id=42791&code=su7300

Here's the RMC registry mod hack.

http://forum.notebookreview.com/6130270-post4855.html

 

So I can only undervolt the max multiplier? Oh well it runs pretty cool anyway, like 19-20C idle

 

Same with me, but it only happens when I'm trying to use the undervolt on battery. Sometimes it would bluescreen in Vista and then each time I restart it would show me the same bluscreen while booting up. Then I would have to either let it stay powered off for a while, or remove the battery and hold the power button for about a minute to get the laptop reset. Just can't understand why the undervolt I have is perfect when plugged in, but bad when on battery.

 

No. You can actually undervolt all integer multipliers and the max multiplier.

 

Okay, I haven't read through all 545 pages of this thread obviously, but I have done a search and found nothing. I undervolted exactly according to this guide - unchecking IDA, which has the highest multiplier for me - and lowered the 10x multiplier (the next highest) by .100.

..my temperatures actually got worse, and my computer overheated and crashed within a couple minutes of running the test. No BSoD - just a warning from RMClock that my cores were overheating then my laptop went into sleep mode.

BTW, my computer is running extremely hot. It idles in the 65-70 range and shoots up to 90+ when playing games - frequently causing overheats and crashes.

How is this even possible?

Any ideas at all as to what to do? I've tried using compressed air to clean out the vents/fan already.

 

You didn't mention what CPU you have but if your CPU core temperature is going over 90C then it would be a good idea to pull your laptop apart and have a look at the heatsink. It is likely making poor contact with the CPU so heat transfer is poor. For a heatsink to work properly it has to be firmly attached to the CPU so it not flopping around. You will have to redo the thermal paste if you take it apart. When the heatsink is not working properly, it's very easy for the core temperatures to go sky high within seconds of applying full load.

 

Thanks, that sounds like that is my problem. I have a Lenovo T61p - apparently there are some known issues with this happening with this particular chip (and the nvida 570m GPU especially) so I guess I shouldn't be too surprised. I've gotten nearly 4 good years of life out of it, though, so I can't complain.

I'll look into how to check/replace my heatsink.

 

Hey Everybody.

New final version of Notebook Hardware Control has been released. The current version number is NHC v2.4.2 and it was released December 14, 2010. Right now it is only 32bit version but 64bit should be ready very soon.

Image


This aplication is great alternative for undervolting.

For more info and download please visit author's web page: Notebook Hardware Control (NHC)

Direct download

 

Just installed win7
Installed RMclock
Trying to do the reg fix for the half multiplier but I cant find the Rightmark folder under USERS>Software for some reason
It was there under vista x64
Any ideas?

 

You should be able to find RMClock in your registry here:

HKEY_CURRENT_USER\Software\RightMark\RMClock

You need to run the program at least once for it to create this registry key.

 

I've been using rmclock for 6 months.

.95v for all speeds 6x SLM to 12x for a t9300 penryn. I'm pretty happy with it. How do I get .5 multipliers?