The "Undervolting" Guide

If you're cpu is on the max multiplier all the time so its no idle at all..never..I don't understand how is that possible btw..? For what I remember Windows 98 was the last OS which used to drain a 100% cpu all the time..

 

If I change the setting to "No management," it goes back up to the highest multiplier. There's no change in CPU usage.

 

Okay, so it is okay to change the lowest multiplier setting??? If I am reading Flipfire's guide correctly, it says to never do this. But from the above comments it appears some people are doing so without harm.

 

What flipfire meant with "cant/shouldn't" is for CPU like mine, where I cannot change the lowest multiplier's voltage, because it is already at the lowest by default, there is no lower voltage for me to choose. flipfire probably tested on a similar CPU when he wrote the guide. But for some other CPUs, if it can accept lower voltage, there is no harm as long as it's stable after tests. Many users have been able to lower their lowest multiplier's voltage without any problem.

Also, your CPU should have a certain hard-coded limit as lowest voltage for the lowest multiplier. For example, I tried the RMClock registry hack to unlock all the VIDs so that I can force to choose something lower, but guess what, even if I choose 0.8500V for my lowest multiplier, the CPU will still run at 0.9000V for idle.

Anyway, just make sure you run some tests, there is no problem lowering the voltage of the lowest multiplier. Many people have done it.

 

Well in that case, the idle temp won`t drop. But the max temp is what is usually the problem anyway...so problem solved

 

Tekito,, i think Stewie is correct... my CPU a lowly T5450 defaults at 1.25V for all multipliers... but after testing i've found that it runs stable at 0.95v across the board (even the lowest multiplier)...
but as stewie points out some CPUs already have a minimum voltage for thier lowest multiplier that is substancially less than the standard voltage at max multiplier... and what flipfire was saying (if i understand correctly) is that intel has already undervolted those CPUs to the best stable setting and RMCLOCK usually won't even offer a lower setting...

i could be way off base... wouldn't be the first time

 

Newer CPU's you cant change the idle voltage, the lowest voltage possible is hardcoded, you wont be able to go any lower. Meaning your idle temps cant be changed than they already are. Unless of course you have SuperLFM.

Pentium M's you were able to drop the default idle voltage from .988v to .700v flat with RMclock. That sure improved my idle temps. My fans doesnt even need to run anymore after i undervolted which pretty awesome.

Your default voltages werent detected properly, your 6x idle voltage should already have been set to 0.95v and going higher and higher for each multiplier till it the max multiplier is 1.25v

 

Okay, many thanks (to everyone) for clearing that up.

Indeed, that would be a dream come true.

 

Official Benchmarks

Tested on a Compaq M2000 using PCMARK05 (System and CPU suite test)
-Pentium M 725 1.6ghz
-768mb RAM
-40gb 4200rpm HDD
-WinXP Pro

Default voltage
.988v - 1.340v

Undervolted:
.700 - .988v

Test results as folllows:

No RMclock:
http://img241.imageshack.us/img241/3826/cpuscorewormclockfl7.jpg

System Points: 1494
Max temp reached: 66c

With RMclock/Undervolted:
http://img401.imageshack.us/img401/3427/cpurmclockvl6.jpg

System Points: 1474
Max temp reached: 55c

-

As you can see, points are marginally the same, meaning only a very very tiny performance loss due to the RMclock program using some resource to power itself.

The test's peak temps dropped by about 10c when undervolted.

I did the test three times just to make sure it was accurate. All tests were run under the same conditions.

 

My max multiplier is on 1.004V at the moment. With 0.988 it crashes every so often when I play on big servers on CS:S... so i had to put it back up again.

 

Thanks for the response Tusin.

Have disabled the soundcard as you suggested, and the CPU load still appears to fluctuate between 40-60% whilst OS load is small at 2-5%. Any other thoughts ? still getting a drop of around 100 on PCMARK05 when undervolting.

When I was testing stability last week i ran Orthos overnight for 9 hours and had a max temp of 70 recorded the next morning. when I ran the same test in the office just after setting the 10x multiplier i got around 62 degrees (missus likes the house quite hot so suspect this is the reason for the higher temp, all temps appear 8 degrees hotter at home !!!!)

 

Don't worry Johnny, I don't think the temp will be that much of a difference between 1.004V and 0.988V.

 

Thanks OP, was up in the 60's now down in the 50's!! Great post...

 

just did my first undervolting on my T7500. 1.1500V was the lowest i could go, the next lower voltage wouldnt last more than a couple of minutes

Results:
first time MAX before undervolting= 87 celsius (10 minutes of stress time)
results after Undervolting = 77 Celcsius (35 minutes of stress time)

So an 18 Degree difference in Fahrenheit, not bad

Here r the numbers for my T7500

7x (800 MHz) = 0.9625V
6x (1.2 Ghz) = 0.9625V
7x (1.4 Ghz) = 0.9750V
8x (1.6 Ghz) = 0.9875V
9x (1.8 Ghz) = 1.1250V
10x (2.0 Ghz) = 1.1375V
11x (2.2 Ghz) = 1.1500V

 

Are you sure you disabled it? As in, right click in device manager and select disable? Sometimes you have to reboot for it to actually disable. I have not ran PCmark05, but in 3dmark06 on the CPU test I lose about 70points with having RMClock running. Maybe it is just the CPU? I am ordering a new one tonight. So we will see.

 

So every T7500 behaves differently.
Running at 1.05 right now and super stable.Max temp 65C with orthos, usually 55C when gaming.

 

very coolguide + rep

 

Mine behaves the same too.
Any lower and it will blue screen.

 

Would this mod be worth it for a Thinkpad X40 with an already low voltage 1.4ghz Pentium M 738?

 

Yes, even on an LV processor. Give it a try

 

Definitely was disabled, and when it looked like it had made no difference, i did reboot and made sure that it was still disabled, again with no clear difference.

Looking at your sign, i see you are XP based whilst i am Vista so i suppose that this is one difference.

Just seems strange doesn't it that the CPU is consistently around 50% when the laptop is really at idle and no software is being used. I suppose some process loaded at start-up must be loading the CPU in this manner ???

 

Benchmark update.

I did another PCmark05 test. This time with speedstep OFF using RMclock. meaning my CPU is locked on full throttle.

http://img248.imageshack.us/img248/9537/ssoffaf5.jpg

System Points: 1519

Say what!? We got 20+ points more with speedstep OFF?

Heres the thing about synthetic benchmarks, the only reason we got higher points with speedstep OFF is because the CPU didnt need to throttle up.

PCmark made the CPU to throttle up and down in between the tests. This is where the points is lost, the .200ms throttle time.

On actual use, like gaming the cpu doesnt actually need to do this. Meaning there really is no performance loss on real life applications. (eg. your FPS will be the same because your CPU is usually throttled to full power when gaming)

This is why syntethic benchmarks does not always accurately represent the real performance. They should only be used with comparing with others who ran the exact test parameters.

 

I guess so, I have mine at 1.000v but is hotter than yours with orthos..68ºC here . I guess the laptop design is a factor too

 

It's not normal that RMClock uses 50% CPU constantly. When you open Task Manager, which process(es) is using the CPU? You can see a list of process and the CPU usage.

I'm pretty sure the laptop design plays a big role too, and also the amount dust in your system.

 

flipfire,

While you're at it, I think you should also test "P-state transition method" and "Multi-CPU load calculation" settings to see if it makes any difference.

 

Yep, I have a 17 incher, and I just performed my monthly dust cleanup

 

No dust, cleaned it up 5 days ago

 

See below list;

ApMsgfwd.exe 236k
ApntEx.exe 496k Alps pointing-device driver for windows
Apoint.exe 720k Alps pointing-device driver
BTTray.exe 924k Bluetooth tray application
Csrss.exe 1328k
Dellwmgr.exe 972k Dell webcam manager application
Dwm.exe 34636k Desktop windows manager
Explorer.exe 23496k Windows explorer
Hidfind 376k Alps pointing-device driver
IAAnotif.exe 312k Event monitor user notification tool
Mcagent.exe 320k McAfee security platform
Mcuimgr.exe 1104k McAfee user interface manager
OEM02mon.exe 600k Live cam console auto launcher
Psqltray.exe 1920k Fingerprint tray application
Quickset.exe 1020k Quickset
Rmclock.exe 1864k Rightmark CPU clock utility
Rundll32.exe 308k Windows host process
Sprtcmd.exe 1124k Sprtcmd
Taskeng.exe 2048k Task scheduler engine
Taskmgr.exe 2176k Windows task manager
Upeksvr.exe 1972k
Winlogon.exe 1188k

Obviously there are 2 sizable items in that list, should they be that high ?

 

Those numbers are the memory usage size, not the CPU %, the CPU % is the column next to it on the left. Which ones are using high & of CPU?

As for the 2 sizable item, don't worry about it, my Firefox process is much higher than that, 133,532K atm. That number is just telling you how much memory is using.

 

This is the memory usage, not the CPU. For explorer.exe is normal to use this amount of RAM, I don't know what is Dwm.exe.

Edit: @stewiegriffin dude, slow down

 

http://www.neuber.com/taskmanager/process/dwm.exe.html

From my avatar/sig, you can't tell I love speed?

 

Apologies, i had already got that list from the weekend when i was playing around and never thought to check the values.

With respect to CPU all values are chown as 00 other than those below;

dwm.exe fluctuates between 00 to 02
explorer.exe fluctuates between 00 to 01
rmclock.exe generally 01 constantly
taskmgr.exe generally 05 to 06 constantly

 

The CPU usage appears very normal.

Which monitor software did you see your CPU fluctuate between 40-60% at idle? It seems like that monitor is not reading the CPU usage properly.

 

Read it using the rmclock monitoring page/taskbar icon.

 

I just checked mine, it seems like RMClock's monitor reads a bit differently than Windows Task Manager, for me it's not as high as 40-60%, but it's about 10-20% higher than what WTM reads as the total CPU usage at the bottom of WTM.

In your WTM, what is the total CPU usage %? It's at the bottom of WTM.

 

Just checked mine, RMclock 60% WTM 10%..So what?

 

Mine roughly matches up with the value of OS load shown on the rmclock monitor, with the CPU load at the higher value i mentioned.

ie WTM is 5-6%, OS load is 5% and CPU load is 52% also ran resource monitor from WTM and this is replicating the WTM as you would expect, and the percentages shown here seem to fit with a 5% usage ?

 

This might need to be discussed in the RightMark forums

I speculate the 60% cpu load corresponds to your current clockspeed. So let say your idling at 1.2ghz, 60% of that perhaps?

My XP notebook gets 1-20% cpu load with rmclock. My vista notebook shows 40-60% on idle. weird inconsistency...Could just be vista screwing around...

dwm.exe is aero btw.

 

I wouldn't worry about it, RMClock's monitor is probably just off. Either that, or M$ is trying to fool us with WTM about how little CPU usage Windows is.

You can also download other free CPU monitor and see the difference.

http://widgets.yahoo.com/widgets/cpu-monitor-1

EDIT:

I think this Yahoo one reads off from WTM, so you probably won't see a difference.

 

You're right. I'm idling with SFLM(600Mhz), but SFLM is configured at 6x (1.2Ghz) so it does make sense. 50-60% cpu load from 1.2Ghz is like 600 - 720Mhz=SFLM?

Edit: Tried different multipliers for SFLT. SFLT remains at 1/2 of the selected multiplier frequency, and CPU load remains 50-60%. So the calculation above works for any SFLM multiplier

Edit2: I disabled SFLM all multipliers but the x11(the highest) and now RMclock reports CPU load exactly like WTM. CPU load in RMclock takes the multiplier which is running in the moment and calculate the percentage of the max multiplier..
In my case: 2200Mhz/100*CPU Load(%)=Current Multiplier. CPU Load=Current Multiplier/(2200Mgh/100).SFLT off.
1200Mhz(6x)/100*CPU Load(%)=SFLM. CPU Load=SFLM/(1200Mhz 6x/100). SFLM on and set at 6x
Issue resolved?

 

thank god for my quad core brain. lol

you shouldnt worry about CPU load anyway. Its supposed to fly up and down. IF you actually start noticing a loss of system responsiveness, then worry.

 

Sorry about that, can't explain it better, poor english
I'm not worried, just tried to figure it out

 

I`ve disabled aero for this exact reason.
It seems to go really bad with Ashampoo(h) antivirus and RMClock for some reason.(High CPU loads)

 

I just received my XPS M1530 not two days ago and I've decided to tackle this "behemoth" of a project, akin "undervolting". Now I understand the time consuming process of testing each voltage at each multiplier yata-yata-yata until the dreaded BSOD, but for me personally, I've never encountered said BSOD. Instead, I get an error message on ORTHOS, so I was wondering if that dictated the same message as BSOD?

And is it true that undervolting does indeed affect performance if one is running a T9300 as I am? I have read that because RM clock does not support a 12.5x multiplier, we 2.5GHz-ers are forced to run at sub-2.4 speeds.

And finally, I've undervolted the 12x multiplier to 0.975, down from the factory setting of 1.1375 (iirc), but it's still running at a scorching hot 74C under load (granted, that's down from 81C at stock voltage), am I doing something wrong here? And if it makes a difference, which it probably does, I'm doing this in the middle of summer in California, where room temp ranges from 85-95 F.

EDIT: I've just read that IDA (13x multiplier) functions as the 12.5x multiplier and by checking the box thing, T9300 users are able to gain access to the 12.5x multipler to undervolt. Can anyone clarify?

 

1) UV does not affect performance,if anything, lower temps are safe calls for all parts.
2)pretty much yes,you could lose 100Mhz, but you`ll need to double check this with T9300 users...
3)your T9300 runs at 74? wow, that`s hot. Penryns run cooler by default,something is weird.You should get lower temps than that,but then again, the XPS 1530 runs hot, due to the GPU.
It has only 1 heatsink as far as I know,so heat is shared between CPU and GPU.

 

Lol, I answered and he deleted the questions

 

Hi

I have merged this thread

If you tick the 13x (IDA) multiplier, this will also use the 12.5x multiplier. The actual IDA feature is a little buggy so i suggest you go to the Advanced CPU options and untick "Engage IDA" > Apply

 

Well, HW Monitor shows the core temperature at 73, but RM clock shows the core temp at 67, which one to believe...

Yeah, the GPU is pretty damn hot. I've never seen it under 60; it's at 74 when Orthos is running.

 

Try using Coretemp as a third program to see which temp is correct.

You get inaccurate readings sometimes due to different BIOS/ACPI's. It will read the wrong sensor.

Also the temp program itself can have a glitch for your CPU

 

Orthos error message in equivalent to BSOD correct?

And also, for some reason, the core clock registers 2.3GHz instead of 2.5 now...

EDIT: Alright, just "finished" testing the 10x multiplier for 30 minutes. It seems to be stable at the 0.9250V minimum. Would it be safe to assume that all of the lower multipliers would run stable at that voltage too?