The "Undervolting" Guide

ya got to figure out a way to lower the lowest voltage
my "idle" temp with a browser and msn on is around 45 degrees, want to lower it if possible

 

same here... my laptop used to be a lap fryer, and i live in florida i don't need one...
but even at .9500 its still about 45ish +/- 3 at idle... i wanted something in the mid to low 30s idle and 40-50's max... i mean these are 45nm penryn's for pete's sake...
they shouldn't be hotter than meroms...

 

what kinda video card do you have? i have a 9500m gs which could be the issue. idle temp is around 45 degrees

 

whatever the hp 14" comes with the 8400m gs?? which i think is a 9300m gs?? same thing?

 

i found this in a different thread:
1) set the FID/VID transition time to 50us (or a bit higher) - this allows time for the processor to stablize before changing VID
2) set the p-state transition method to "perform single step transition only" - this makes sure that the transition between different FID/VID's will not change so significantly that it destablizes your computer


credit: r34p3rex

i applied the settings haven't uv'ed again...

 

i get this:
Test 1, 4000 Lucas-Lehmer iterations of M19922945 using 1024K FFT length.
FATAL ERROR: Rounding was 0.4921875, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 0 minutes 58 seconds - 1 errors, 0 warnings.
Execution halted.

does that mean i can't go lower??


this was in orthos...

 

HWMonitor said my ACPI sensor hit over 6000 F, so the program (Or the sensor itself) clearly glitches at times.

 

I'm idling at 45 even with a 9500 GS

Not exactly, the 9xxx series cards are the same basic core, but they are smaller and produce less heat

 

Now that looks much better and normal. When you have free time, do another test of your highest multiplier for like an hour or more. But I think it should be fine now.

 

oops i meant to type 45 degrees on idle

 

Here's my final stable T7500 after testing all multipliers. I didn't need to test below 9x because it already runs at my lowest available voltage.



For this CPU, that appears to be a very good undervolt. Everything but the highest multiplier stays below 60C. Also, does 1.8 GHz @ .9v run inherently hotter than .6 GHz @ .9v? I'm idling at 45-49C at 1.8 Ghz which is just about what it ran at with SuperLFM checked. Given that the temps are just about the same, I unticked 6-8x in my 'Performance on Demand' profile since 9x runs at .9v so there's really no need to even use the lower multipliers on AC. My idle is 1.8 GHz so my base line performance is higher. Any downsides to this? For the other T7500 people (Difference systems than me, but still curious), what are you idling at and what are your multiplier/voltages at?

 

I don't have the exact CPU, but I didn't have to test the lowest either, cause it cannot go any lower than 0.9000V. Also, same here, my idle temp didn't really dropped much, only a couple of degrees compared to the default factory setting, but I still enabled SLFM and set it at 6x which makes my CPU idles at 600MHz.

Have you tried making it idle at 600MHz? The idle temp didn't drop by a couple of degrees?

Anyway, I don't see any problem idling at 1.8GHz if the temp is pretty much the same. Although I still did it for 2 degrees.

 

With SuperLFM when actually using my computer, lowest I ever see is ~ 45 C. I run HWMonitor though all the time to get an idea for minimum temps, and lowest I've seen is 39C, but any time I'm actually on it, even with 600 MHz SuperLFM, it's 45-49. Which is just what I'm getting at 1.80 GHz.

In terms of battery life though, since 600 Mhz and 1.80 GHz are running at .9v, will the higher MHz figure make the battery life shorter or will it be the same since they're both taking the same amount of power. If I can run at 1.80 Ghz without taking a battery life hit, that'd be pretty sweet.

 

Both are set @ 0.9V, they will use the same amount of battery only if both have the same amperage. I believe 1.8GHz will definitely have a higher amperage, probably not 3x as much, but definitely more.

Someone corrects me if I'm wrong.

 

I actually couldn't see settings below 0.95V, either, once I followed the Advanced CPU settings of the OP.

Here are my settings:
6.0x = 0.9500V (SuperLFM)
6.0x = 0.9500V
7.0x = 0.9625V
8.0x = 0.9750V
9.0x = 0.9875V
10.0x = 0.9875V
11.0x = 1.0000V
12.0x = 1.0125V

It dropped the load temp from 64C to 50C, and I think it lowered the idle temp by a couple of degrees. It usually idles at 22C now.

 

Btw, does anyone know the recommended settings for "P-States transitions method" and "Multi-CPU load calculation"?

 

damn... 50/22 is nice... i am not getting that... i have a 8400m gs which i dunno if it shares the same heatpipe as the cpu could that be causing higher temps? what laptop/specs do u have xeci?

 

I use both the powersaver and RMClock and got the best battery life yet(2 h at full brightness),but with SLFM at 6x.

Not sure I wollow, the P-states usage means the CPU only runs at the frequency needed to get the job done,not faster.
I use all P-States on ACmode and only the 0(6x in my case)multiplier for battery...
600 mhz dual core is enough for browsing and such.

 

does this happen to anyone?

some times when i have rmclock running my voltages jump from custom (ie 12x 1.0125) back to the orginal (12x 1.135)

 

when you configure vista powersaver, do you select rmclock as your power plan?

 

Hi
I only have 6x and 7x for my E6550.This limits this undervolting a lot right?The 6x is at the lowest voltage of 1.185 and there is no value lower than this. For 7x,I lowered it from 1.3xx to 1.200.

Much lower temps from 58-45.

Thanks guys.

 

Well, you have a desktop CPU, so that works a bit differently,but dropping to 45C is a nice one...

 

As I thought .Thnx Eleron

 

Ok, strange little problem here.

As directed, I enable PST in Profiles>Performance On Demand, and check each of the indexes. Once I go back to Profiles, and select Performance on Demand, my CPU steps down as one would expect.

When I start up Orthos, however, my CPU scales up to a 96% max load (and 100% OS load). This 4% throttle then remains active even if I disable PST on that profile in RMClock, or switch to a different profile without PST enabled. Even if I exit RMClock, that 4% throttle stays in effect, costing me approximately 100mhz.

The only way I've found to remove the throttle is to restart, or as it turns out, switch to a different power scheme in Windows XP and then back to my normal profile.

Frusterating, as I don't want a permenant 4% performance loss =/

Edit: Ok, changing the windows power scheme doesn't always fix the problem, apparently, just some times. Any idea why I'm getting throttled?

 

the T9300 has a 12.5x multiplier. RMclock can only recognize whole multipliers (11,12x) not half multipliers. The reason its throttling back is because 12.5x is not active. The highest multiplier active is 12x (2.4ghz).

You can either turn IDA 13x multiplier on (which will also activate 12.5x multiplier) or simply overclock for the lost 100mhz. IDA is sort of buggy and might cause BSOD's even if you lower your voltages a little.

Try the following first:

1)Go to the Advanced CPU tab > untick the 'Engage IDA' box > click apply

2) Go to the main profile page > Tick IDA > Apply

3) Go to PoD sub-profile page > Tick the Index for 13x IDA > Apply

4) Try and lower the voltages for 13x and hope it doesnt bsod when testing

 

Righto. Will lowering the 13x voltages lower the 12.5 voltages as well?

 

Yes the 13x will act as the 12.5x

Let me know how it goes...

 

has anyone figured out how to lower the min voltage?

the lowest voltage i can set is 0.9500 for my t9300, anyway to set it lower than this?

 

For P-states transitions setting, I believe "multi-step" is better, which is default already. And for load calculation, I just left it as default also, the default SUM setting is recommended by RMClock.

This is the detail description of them::

 

I read that you can hack the registry of RMClock and make it lower than Intel's recommendation, but I haven't tried that myself yet. I might try it later since I don't believe 600MHz requires 0.9000V, plus other people were able to go to 0.8500V, although I know it's not the same for every CPU. If you try it, let me know how it goes.

http://www.mouldy.org/extending-battery-life-on-the-dell-xps-m1210

I don't recommend you to just run "RMClock_Tweaks.reg" because there are many settings in it that you probably don't need. I opened that file and to unlock all the VIDs, all you need is the following:

 

Tried a lower voltage but it won't actually apply, even if I choose it.

 

You mean you just tried to unlock all the VIDs in registry and select a lower one than factory default?

 

I tried it weeks ago actually. .9000v is the lowest my CPU goes -- the registry tweak unlocks it down to .7125 (or something) and I can select lower than .9000v, I can hit apply, and the program says it applies, but the CPU info tab still shows it as .9000v, as does CPU-Z. So my guess is that it's hardcoded into the CPU somehow.

 

Same here, I just tried it, can't go lower than 0.9000V also.

 

According to intel, the penryn T9300 voltages are actually:

"VID Voltage Range: 1.00V-1.250V"

Your lucky enough getting .950v as your lowest voltage. Dont bother trying to tweak it. Im not suprised a T9300's minimal voltage is .950v

Source:
http://processorfinder.intel.com/DetailsPrinterFriendly.aspx?sSpec=SLAZB

 

Your advice worked out perfectly.

Running my T9300 OC'd at 2.8ghz at the stock voltage (1.2250v) I hit 59c.
I got it down to 1.1000v so far, with a peak of 54c.

Not as dramatic a heat saving as some previous folks, but I was starting out from a much lower temp. I'm definitely happy about that, though - it's more than a 10% reduction in voltage at peak, and a similar reduction in processor temperature. Best of all, my fans run between low and medium, not once hitting high. I've no idea what impact that will have on my battery life, but it certainly can't hurt!

Thanks for your help!

 

I have Clevo M570RU with 8800M GTX. My CPU temps seem to be lower than those of the same specs, but my GPU idle temps seem to be a few degrees higher than the typical.

 

Regarding 2), I've also read some guide recommending it to set to "Perform single-step transitions only" rather than default "Allow direct multi-step transitions." This is under the "Management" tab, btw. So does anyone know which one is better?

 

I had this option set to multi-step transitions on my old Latitude D610. Worked with no problems for 2+ years. I think this option is more of an extra stability measure than anything else.

 

1 step : from 600 Mhz to full speed if required
more steps : graduatly increasing if needed

The way I see it, multi step is the better choice.

 

So in short a processor with more multipliers are better?

 

Yes, since not all apps require the full force. Divide et impera

 

I should look for that from next time.

 

I saw someone mention the registery tweaks to unlock lower VID's. I understand it's kinda hit or miss, but if I tried that to get lower than the minimum 1.0375 available, is the worse that could happen still just BSOD? If so, could someone explain how to put that setting into the registery? I haven't done manual edits of it in the past and not sure if it's just a cut&paste thing or what. If it's something fairly complicated, feel free to tell me it's not worth the bother.

 

It's fairly simple to edit the registry:

Start -> Run -> regedit

Then browse to [HKEY_CURRENT_USER\Software\RightMark\RMClock]

On the right panel, right-click anywhere and select New -> DWORD value

Then type " UnlockVid" as name without the "

Then double-click on it and enter 1 as decimal value.

That's all.

After you make any changes, you must right-click on the RMClock icon and select "Restart RightMark CPU Clock Utility" to see the effects.

You can always delete that DWORD value later if you don't find it useful for your CPU, or just leave it as 0 which is disabled.

 

Awsome, thanks.

*goes off to tinker some more*

*EDIT* Seems the E8400 won't accept the lower settings. Oh well, I tried and thanks again anyway.

 

While working down the line to the lower multiples, I discovered something interesting. It seems that my processor never goes below 1v, though it lists the 6x-9x multipliers as all set to voltages below 1. Even at those multipliers, however, the CPU info page shows a current voltage of 1v.

Is this normal for the T9300?

 

This is normal, the CPU will not always accept RMClock's setting, I believe the lowest setting is like hard-coded in the CPU. For example, I can't make mine runs under 0.9V even if I force it lower, it will still run at 0.9V.

Another thing I noticed is that it might be influenced by the bios/motherboard because I've seen people with very different settings with the same CPU, but they don't have the same laptop though.

 

Fair enough. I was just surprised as those are the settings that RMClock started out with - I assumed they were the CPU defaults.

Actually, they probably are the CPU defaults and you're right - the motherboard/bios have their own limits as well. Nevertheless, I'm fine with that - I got a fairly significant drop off the high end multipliers, and that's where the real heat is generated anyways.

 

I ran into a problem.. Might be simple. Everything works fine, Voltages read at what they are set at.. Until I check IDA.. Then..When I run the stress test, The cpu is running at default voltage in the cpu tab. Unchecked, And my cpu is running at my set voltages again.

So, Do I have to run without IDA? Or is there something I missed?

Thanks!