The "Undervolting" Guide

It looks like you missed this page in the under volting guide.

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

 

came across this site and this is my first post, hopefully someone can be of some help.

so i'm in the middle of trying this and after changing my first value in the highest multiplier and starting a stability test (4th test, 16 minutes and counting), the hwmonitor is showing that my core temperatures (originally posted at 95 and 92 degrees, ridiculously hot for some reason) have actually INCREASED and maxed out at 100 degrees celcius.

i am also getting a noti...

ok in the middle of typing that, the computer shut itself down. no blue screen, no errors, just turned off. surprisingly, and to my delight, once i rebooted, google chrome actually went as far as to restore this open site AND what i was in the middle of typing.

so back to this, a few questions:

1. am i doing something wrong?
2. is my laptop supposed to get that hot to begin with??
3. why, when i lowered the voltage, did the temps increase?

i hope someone can help me out because a: this laptop heat is getting to be an annoyance more so than a noticeable performance killer, and b: because i would like to have some type of extended battery life and/or performance increase.

thanks for the help in advance, i hope to hear back!

(i guess i should include some things about the notebook eh?)
HP dv5-1235dx
intel core 2 duo
t6400@2 ghz
4gb of ram
windows vista 64bit

 

ok so i still want to make this work somehow, but figured i would try an alternate core temp program first.

using coretemp64

@5% workload
core0 - 43 degrees
core1 - 43 degrees

@100% workload using orthos
core0 - 85 degrees
core1 - 84 degrees

not a huge difference, but if the second program(coretemp64) is more accurate, i'll take what i can get i did notice from reading through the first few pages of this thread that someone else had an accuracy issue and i believe they were using vista as well. either way, onward i forge to the stress test! if anyone has any helpful advice, i'd appreciate it!

 

so again, even after voltages are DECREASED the temperature INCREASED. i'd love some insight from some of you smart folks

the computer was about to shut off again so i stopped the "small ffts - stress cpu" test @ 5 minutes 14 seconds

core0 reached 90 degrees
core1 reached 89 degrees

again i started with the highest multiplier (10.0x) and lowered the voltage from 1.2625v to 1.2500v.

i want to try going lower, but if the temperature is only increasing, then should i even try??

signed,
my lap is hot, need help

 

ok i was curious as to what would happen (and i just REALLY want this to work) so i dropped the voltage this time (again on the 10.0x multiplier) from 1.2500v to 1.225v and almost identical reaction:

@ 5 minutes 22 seconds again both cores reached 90degrees and windows was about to shut down.

is this rightmark program even doing anything?? 3 voltage changes and no change in result. 45 minute stress test?? lmao this won't even get to 6 minutes!

sorry to be repetitive but if anyone is familiar with why this is happeneing, or can offer any help, i'd really appreciate it

 

Here are the Intel specs for a T6400.

Intel® Core 2 Duo Processor T6400 (2M Cache, 2.00 GHz, 800 MHz FSB)

Make sure HWMonitor is using the correct TJMax value which is 105C for your CPU. You usually need to edit the hwmonitorw.INI configuration file to fix this.

CPU_0_TJMAX=105.0

Also make sure Core Temp is using this value. You can adjust it in Core Temp if it is not.

If your laptop is hitting 100C during a simple stress test then it has some serious problems. Here's my T8100 running at 2.1 GHz for half an hour at default voltage.



Furmark is also running, heating up the GPU and the entire laptop but the CPU core temperature is only 63C /64C. A 100C core temperature means either HP doesn't know how to design a heat sink or it is clogged with dust and dirt or it's loose or you live outside in Phoenix and it's July. If it was my laptop, I would pull it apart and make sure that it is clean and tightly attached to the CPU. A poorly designed and loose fitting heatsink is not unusual. You may have to do some re-engineering with washers or what have you to tighten it up.

Once that is done, then you can get back to under volting. A difference of 0.025 volts is insignificant. A half decent mobile Core 2 CPU should be able to run 2.0 GHz, 100% reliably, with 1.05 volts. A good CPU might run that speed with 1.00 volts and some users with P series CPUs are running reliably with less voltage than that. With your CPU, I'd start testing at about 1.10 volts and see how things go and then drop it down in steps of 0.025v. Under volting can only do so much if your heatsink is loose or dirty.

One user today concluded that running Prime95 is a better stress test than Orthos. Prime95 has been updated a few times since Orthos was released.

Prime95

 

hey thanks for the reply first off!

so secondly, without even checking coretemp64, i remember seeing that the tj max (where i bought some pants ) was set at 90 so that might be why it keeps shutting off and i'll try raising that to 105

thirdly, even if i change that, like you said, 100 degrees (and anything remotely close to that) is ridonkulous, BUT i have NEVER had ANY hardware or software malfunctions, even using vista. the machine has run smoothly since the day i bought it with the only hardware change being a 6 cell battery over the mammoth annoying 12 cell. the only reason i came across this site in the first place is because i was bored today looking for some tweaks

fourthly, opening up the laptop is beyond me. i'm no big dummy (sanford and son ), but that is an endeavor i think i should leave in someone else's capable hands if i don't just drop this undervolting adventure altogether

fifthly, as far as living conditions go, i live in the middle of austin, texas (god i miss new orleans, but it would probably be equally as bad for a computer), but it's been chilly so the outdoor conditions wouldn't play a part in this. the indoor conditions however can get pretty warm in our house and dust definitely builds up, i'll have to get some air duster at the very least and do what i can

sixthly, i rarely have the battery profile set to high performance so that type of temperature is...well...rare, but as aforementioned, and extended battery life would be nice.

seventh, is there any way around opening up the laptop? i had an unfortunate experience with my previous laptop (compaq v2000) where i replaced the lcd inverter and smoke poured out of the vents. i don't necessarily think i did anything wrong, but my point is i don't have the best of luck opening up machinery.

again, thanks for your reply, and any further help is much appreciated!

 

This thread is so huge.. What about AMD cpus? Is it possible to undervolt AMD P520? If yes, how?

 

If RM Clock does not work with your AMD CPU then try K10Stat.

Setting TJMax correctly won't change your actual core temperature. It simply corrects the reporting of your core temperature. Here is Intel's formula:

Reported Temperature = TJMax - Digital Thermal Sensor reading

The sensor works like a reverse thermometer. It counts down as the core temperature goes up. That formula converts the sensor data to an approximate temperature. When you use the correct TJMax, your reported core temperature is going to be even higher.

Intel CPUs should be able to run reliably right up to the TJMax temperature. If a heatsink and fan are designed properly, that heat will be exhausted away from your computer so you can comfortably use your laptop on your lap.

If you're not comfortable taking your laptop apart, you can still try to do some under volting. This will help if you can get your voltage low enough but it's probably only going to be a partial cure to the real problem which is too much heat.

 

by SLM do you mean SuperLFM? does that mean you changed the SuperLFM to 12x? I thought that was meant for low speeds?

 

I think he meant that he can run any multiplier, up to and including the 12X multiplier with the lowest SLFM voltage which is usually 0.95 volts.

jedisurfer1: A T9300 supports a 13.5 IDA multiplier and a 12.5 HFM multiplier so post a screen shot if you get those working. RM Clock can not correctly monitor these because of the half multipliers.

 

Thanks to everyone who contributed to this guide.

I could successful under volt the P9700 in my E6400 to 0.925 at 6x and 1.0875 at 11x stable using this guide and the registry hack posted to support half multipliers. It now idles close to 30C and remains around 70C under full load tested using Orthos for about 50 minutes...

 

Hi All,

I have an Acer Extensa 5630 with a T3200 that runs as hot as a firecracker. Someone else mentioned their T3200 didn't have IDA or SuperLFM. Mine doesn't seem to have it either.

The default voltage for the 12x is 1.250. I lowered it to 1.2250 and it BSOD'd in less than 5 minutes. If my math is correct, that's only 25mV. Should I be thinking there is something wrong with the CPU, or can these things vary that much?

Anyway, I disable the 12x multiplier and start testing the 11x. I'm currently down to .9875 and it's handling the load fine. It does get up to 85 degrees, but at the default voltage it would hit 92 in about 2 minutes, so 85 is an improvement. I'll keep going and see how low I can go with it.

I would appreciate any input as to why I would BSOD after dropping the voltage on the 12x by only .025.

Thanks
John

 

that's the wonder of semicon. There's nothing you can do about the voltage anymore and seems you already lowered your cpu voltage to it's lowest operating range.

 

A general question about undervolting.
Is it possible that undervolting affect the correct audio reproduction?
I ask because, after some test with undervolting, my audio section had some problems, like crackling or similar...
If i stop the undervolt i have no problem...

 

is there any other tool beside K10stat for for AMD users?

i've got a AMD Turion x2 ZM-82.

 

thanks for the guide
my Acer Ferrari One temp dropped from 53C to 46C on idle, and doing stuff like browsing the net certainly drop to 50C from the 55C before

 

I believe that's the only software to undervolt for that CPU. Why you don't want to use it? That software is great. I even use it on my desktop Phenom II X4 955BE

 

Hi,
I just changed my processor from T5550 to T8100 - with half multipliers. Try to use above mentioned link - http://forum.notebookreview.com/6130270-post4855.html
but Registry looks different for me. Probably because of processor version I have 6 speed steps instead of 4 in example. Than my registry has:

TITLE: PStatesData
Type: REG_BINARY
Content:
00000000 06 0f 01 31 06 0f 01 20 - 07 17 01 20 08 17 01 20 ...1... ... ...
00000010 09 16 01 20 0a 16 01 20 - 0b 1e 01 22 .. ... ..."

Does enybody know where to do this change?

 

In the second line, 0a is your HFM multiplier. a in hexadecimal is equivalent to 10 decimal so it is using the 10 multiplier. A T8100 has a default HFM multiplier of 10.5 so to fix this, change that value to 4A. The leading 4 stands for the half multiplier so when combined with the A, you will get the 10.5 multiplier.

The IDA multiplier is the next one in this sequence. Go over to the digits 0b. b in hex is 11 in decimal. A T8100 supports a 11.5 multiplier so to convert this change it to 4B. That's all there is to it. Now you can use RM Clock and access your half multipliers but RM Clock still doesn't fully understand what a half multiplier is all about so many of the graphs and monitoring data will not be correct.

To test if IDA mode is working correctly, try running a single thread of Prime95 and then use the Task Manager Set Affinity... function and lock Prime95 so it only runs on a single core. That will allow the second core to automatically go to sleep so the core running Prime95 will be able to spend the majority of its time using the 11.5 IDA multiplier. ThrottleStop in Monitoring mode is very accurate at reporting the average multiplier for each core. Try this test before and after the registry fix and you should see a difference in your average multiplier.



Follow the other directions on that web site link you posted. The part about adding SkipCPUIDCheck and setting it to 1.

 

Thanks for this insanely awesome review! And raak for linking me

I have some questions though.

I'm using a HP HDX 16 with Intel Centrino 2 (or so the sticker says) with Windows 7 Ultimate x64.

Under profiles-->performance on demand, where I tick use P state transitions, I only have 2 indexes (0 @ 6x, 1 @ 7x). Is that normal or did I screw something up?

I also lowered the voltage for the 7x from 1.1375V to 1.0V and ran stress test for 45 minutes and everything is fine, I cant go any lower for the volts so am I done now?

Thanks!

 

What CPU model do you have?

If you can enable the Dynamic FSB Frequency Switching (DFFS) option in RM Clock, you can usually access some voltages less than 1.0 volts.

 

I have HP HDX 16 (X16-1354CA).

After I rebooted my laptop running at the new 1.0 volts for the 7x multiplier, the computer was running slow as hell, did I do something wrong or did I pick a bad volt setting?

 

Run a program like CPU-Z so I can at least see what CPU model you are using. I assume it is a P7350.

What type of benchmark testing did you do, before and after, to determine that your computer is running slow as hell?

RM Clock does not correctly support the P7350 because it uses half multipliers. The default multiplier for a P7350 is 7.5 which you can no longer access unless you edit the registry of RM Clock. That is explained a couple of posts above.

 

Just did my old ThinkPad. I'm pretty happy, went from 1.3400v and 89c @ full load to this:



Any lower and it threw errors running orthos, OS was still responsive though.

Gonna do my i3 lappy next

 

Hello, I'm trying to figure out if there's a way to underclock my CPU for my server. I've installed RM Clock, which has allowed me to undervolt by about .25v. However, I'd like to underclock by about 90-95% and as far as I can tell, RM can only throttle the CPU down to 56%.

I just took a look at ThrottleStop. I like that it runs with very little resources, but it doesn't look like I can throttle the CPU with this software.

Am I missing something? Is there a way to throttle down to 5% with either of these?


Thanks

 

forkenbrock: What CPU are you using? You can use RM Clock or ThrottleStop and reduce the multiplier to the minimum and you can also use Clock Modulation throttling to slow your CPU down to a crawl. Depending on what CPU you have, doing this may not minimize power consumption if that is your ultimate goal.

In the Core 2 based CPUs, the minimum multiplier physically possible is 6.0 so you can't reduce the total CPU MHz by 95%. Clock modulation throttling causes the CPU to skip cycles which can reduce performance and power consumption significantly when a CPU is at full load but when a CPU is lightly loaded, this can have the opposite effect and cause the CPU to become less efficient and it will consume more power. Let me know what CPU you have and I'll see if I can do some testing with a Kill-a-Watt meter with a similar CPU.

 

Is there a way to undervolt i3's? I tried RMClock and ThrottleStop, but there isn't any option to change the voltage.

 

Hello UncleWebb, here is my CPU. It's an e6300 dual core.

Intel® Pentium® Processor E6300 (2M Cache, 2.80 GHz, 1066 MHz FSB)with SPEC Code(s)SLGU9, SLGW2

I have a friend who told me he was able to underclock the same chip in his machine by 95%, but unlike me, he has a motherboard that supports underclocking in the BIOS.


Cheers

 

Not all friends are honest. Maybe ask him to send you a screen shot of this 95% underclock because that much of an under clock is pretty much impossible.

A 95% under clock means that your CPU would be running at 5% of its rated 2800 MHz speed so that is 140 MHz. Divide that by the default minimum multiplier of 6 and you have a bus speed of a hair over 23 MHz which is quite a bit lower than the rated bus speed of 266 MHz. Many of the mobile CPUs support Super Low Frequency Mode (SLFM) which means the bus speed can be cut in half when idle but your desktop CPU doesn't support that feature.

The 45nm Core 2 CPUs like you are using are designed to automatically go into low power sleep states when idle and conserve power that way. Slowing them down as much as possible saves next to nothing.

Here's a test I did that shows this.

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

You can use either RM Clock or ThrottleStop and lower your core voltage and your multiplier down to 6. You would need to use a program like SetFSB to lower your bus speed if your motherboard is supported but it's very unlikely that you will be able to get it significantly lower before losing stability. I've never seen anyone get their bus speed down to 23 MHz. I've seen some monitoring software report some silly numbers but that doesn't mean the CPU was actually running at that speed.

emc_2: Unfortunately, Intel removed the voltage information from the MSR 0x199 register when they switched from Core 2 to Core i so there is no multi-purpose software that can lower the voltage on Core i3. Some of the desktop motherboards have software available to lower the core voltage on specific board like Asus Turbo V can do but I haven't seen any software that targets the Core i CPUs.

 

Dang. So I guess that i7's can't undervolted then either.

 

Wow nice knowlegde of HP's UncleWebb, I am running a P7350 :O

I didn't use any benchmarking tools, I merely ran the stress test at 1.0v for the highest multiplier shown (which was 7 not 7.5). After I rebooted everything was slow, like if i clicked the start menu button it would show up like 10 seconds later.

I'll look for the post on editing the registry though thanks!

 

Nope. Intel got rid of simple under volting when they switched to the Core i design.

Google was my friend.

As mentioned, RM Clock does not fully support the newer Core 2 processors so unless you edit the registry, it doesn't have the ability to access any of the half multipliers. The slow menus sounds like something else was going on. When a CPU is not properly supported by a program, anything could be going on.

 

Alright, I will look into the registry editing thing, do you happen to have a guide on that as well, I don't want to mess around in the registry without knowing what I'm doing lol I assume there would be some pretty bad consequences.

I went into the registry and this is what it says in my pstatesdata binary value:

0000 06 17 01 20 07 17 01 20 ... ...
0008

 

Went from 1.1500v and 78c @ full load to this:



Had to use ThrottleStop, since RMClock wouldn't recognize my CPU. Dropped the VID as low as it would go.

 

Nice work emc_2. Not everyone knows that ThrottleStop can be used for under volting Core 2 processors and it also works with the newer 45nm Core 2 CPUs and has full half multiplier support. Pretty hard to argue with those results.

danasht: The numbers in the RM Clock registry that you are looking at are broken up into groups of 4 for each multiplier. The first digit is the multiplier so the first multi in your registry picture is 06 which is the 6 multiplier. If you go over 4 positions you will see the next multiplier. That starts with a 07 so it is the data for the 7 multiplier. A leading 4 is the Intel Core 2 secret code so it will use the 0.5 multiplier. To convert 7.0 to 7.5, all you need to do is to convert 07 to 47 in the registry and that's it. You can follow this guide posted earlier in this forum.

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

 

thanks unclewebb,

after changing the reg for pstates from 07 to 47 it showed up in RMclock all funny like expected (71x) but still didnt allow me to go lower than a voltage of 1V?

Edit:

I just noticed under my battery power options (High performance, balanced, powersaver) there is a new RM Clock Power management section, should i select that as my battery option?

 

Thanks for the guide, I always use to when installing a new OS.

Only read back a few pages, but is there any reason why with XP and Vista i can get below 0.9350, while with Windows 7 thats the minimum it will let me get to. 0.9350 is the maximum Im needing with my core 2 duo.

edit - My laptop froze but no bsod. Would i need higher voltage for 7? seems odd...

nvm, the lower voltages finally appeared

 

danasht: Did you enable DFFS, dynamic front side bus frequency switching, in RM Clock yet? You need that for voltages lower than 1.00 to work. That's on the Advanced CPU Settings page of RM Clock but isn't adjustable on all motherboards or on all CPUs.

 

Nope the DFFS switch is greyed out and I can't access it. I noticed you said that you can use throttle stop to undervolt a computer like mine, will your guide be enough to point me in the right direction?

Also if I underclock/undervolt my Nvidia graphics card using MSI afterburner is that safe and will it lower my temperatures?

 

If DFFS is grayed out, that explains why you can't access voltages lower than 1.00. Try the SLFM option in ThrottleStop. It does the same as DFFS.

I haven't written an under volting guide for TS yet. There's not much to it. Start it, click on Turn On, click on Set Multiplier, lower your voltage and you are done. Click on SLFM too if you need that. Head to the TS thread in my sig if you have any questions. If SLFM is also grayed out then I guess your CPU doesn't support that option.

 

Thanks Uncle, Throttlestop is much simpler to use.

I enabled SLFM because it was available but I don't know what it does exactly.

I was able to turn down all my multipliers to 0.925v (lowest possible) from 1.138v. Ran stress test and everything seemed fine, also CPU-z shows that it is clocked at a voltage of 0.925.

Is there any way to just make the voltage permanent?

Also since I've lowered this voltage can I overclock my CPU from 2Ghz to something higher using throttlestop?

Thank you so much you've been a huge help!

 

When you can enable DFFS in RM Clock or SLFM in ThrottleStop, those options let you access your lowest possible voltage. CPU-Z confirms that you can run a voltage less than 1.00 V now.

I'm not sure what you mean by that question. If you use ThrottleStop and set the voltage adjuster to 0.9250 V then that is as permanent as you can get. Head into the Control Panel -> Power Options and set the Minimum Processor state to 100%. This combined with ThrottleStop will keep your core votlage at that level, all the time. With your CPU, you don't need to get too complicated to reduce heat and minimize power consumption. I don't know of any way to lock that voltage into your CPU unless you are using ThrottleStop or RM Clock.

If your laptop supports SetFSB, you can use that to overclock it but you won't be able to do any overclocking with ThrottleStop. TS can only be used to overclock the Extreme CPUs because they have an unlocked multiplier.

 

Should I set the battery from 5% to 100% as well for the Minimum power state?

 

danasht: Try doing your own testing as outlined here.

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

On my 45nm T8100 Core 2 CPU, there wasn't any significant difference in power consumption.

 

Lol im trying to undervolt my t9900 but everytime i run the stress test my laptop shuts down (this is when i havent even touched the voltage settings.. i just ran it to see what the test is)

 

Hi, I'm sorry if this has been asked before, but I have installed RMclock but there is no "Advanced CPU Settings" tab. Is something wrong?

 

hamzie: Not all laptops have adequate cooling to survive a stress test. If you haven't pulled your laptop apart and cleaned out the dust balls lately then now would be a good time. Core 2 Duo CPUs can randomly shut down when they are running hot.

MIRAA: RM Clock has been out of development for 3 years. You didn't post what laptop you have or what hardware is inside it but it is likely not supported by RM Clock if the Advanced CPU Settings tab is not being displayed.

 

New processor.. new thermal paste.. completely cleaned out the dusts

still happens
Sucks so hard i really want to undervolt

 

So im running the stress test and it says
cpu core overheat detected
current cpu throttle level 95.66% and it fluctuates up and down (is this ...normal?)
i cant tell the real temps because my acer 6935g has a problem and it the temp stays constant from one reboot to another.


Can i just select a voltage and instead of running the stress test just use my laptop?

See if it shuts down or get the bsod?