The Ultimate Pentium-m Pinmod/Voltmod Guide

PENTIUM M HARDMOD DOWNVOLTING

I would like to downVolt pentium m 760 . 0.75/1.1 Would be nice. If possible someone could say with a picture which pins to connect or pend.

The mod is for lifebook s7020 , I have upgraded the ram to 4gb and also bought a 80gb ssd and a new cooler and new battery. Heat has been always a issue. I have some arctic silver and a simmer of copper this time but also would like to downVolt. (I already have downvolted with software but i want hardmod)


(Motherlanguage estonian,sorry if mispelled or.......)

Interesting... I have never been asked this question before. I had never considered a hardware downvolt because I consider software based downvolting to be superior in almost every way.

Downvolting with software will give you a larger range of voltages to work with, allow you to dynamically change voltages and processor speeds on the fly based upon load, and a software based downvolt wil not have any stability risks a hardware downvolt would. Whenever you hardware mod the voltages on the Pentium-m, the processor becomes unable to request some voltages from the motherboard (I try to explain why this happens in post number 3).

Because the mod you would like to do downvolts the processor instead of upvolting, you will have to be very careful because in certian situations the processor could receive a voltage that is lower than what it requests. For example, lets say that at 1.6ghz the processor requests .94v from the motherboard, lets say... at this speed the processor is stable at .908v. Because you have a downvolt mod, the motherboard does not see a request for .94v but instead a request for .812v. The motherboard then supplies .812v to the processor which is not enough voltage for the processor to be stable and windows bluescreens. However had the processor requested .956v, the motherboard would have been able to supply that voltage and there would have been no stability problem. This is just a hypothetical example but it is possible for a situation like this to occur. To ensure this situation never happens, you would need to research every single voltage possibility, calculate which voltages the motherboard is unable to send to the processor and then regulate the voltages the processor requests with a software program such as Notebook Hardware Control. Essentially, Even if you went with a hardware mod, you still may need software to ensure the processor never requests the wrong voltage.

If you have not already, I would check out Notebook Hardware Control software. It is very customizable and allow you to dynamically adjust voltages and processor speed based upon processor load.
Check out Notebook Hardware Control here...
http://www.pbus-167.com/

If you still want to do a hardware downvolt then let me know. A hardware downvolt will require a different kind of pinmod which I do not have any personal experience with, however I did research the subject and I know some members here that might be able to assist you.

A hardware undervolt would indeed be a rather strange idea and a likely cause of CPU unstability, especially on SpeedStep-capable CPUs like Dothans.

Overclockers increase Voltage with Voltmods to ensure stability at higher frequencies and, whenever possible, undervolt through software such as NHC or RMClock to control heat and/or increase battery life on laptops. I can't even imagine the effort and complication required during testing where you'd have to open up your computer, plug/unplug your CPU and place/move the wire bridge and repeat for the various possible frequencies of the CPU.

A hardware undervolt would most likely prevent to reach specific frequencies, increasing risks of system crash or even prevent to lower CPU voltages to desired values, thereby leading to higher heat levels than initially desired.

Forget it and stick to software undervolt!

I've got 3 laptops with the M processor which I would like to mod, what's the chance of getting the picture links to work again on the subject?

Hello everyone - This is my first post and I figured I would help out in regards to the long lost url of http://www2.ijib.com as noted by ozzlo.

Courtesy of the wayback machine, the following url seems to contain the desired info:

https://web.archive.org/web/20060904173728/http://www2.ijib.com:1337/phpbb/viewtopic.php?t=3

and just in case something happens I went ahead and copied the text from the same archive.org url:

/*

Message
ziddey
Site Admin


Joined: 05 Mar 2005
Posts: 3

PostPosted: Tue Apr 26, 2005 2:40 pm Post subject: Howto: Pentium M Overclocking (BSEL0 wire-mod) Reply with quote
Abstract:

The 915 chipset, used in many new laptops, is capable of running at the faster front side bus (FSB) of 533MHz. Pentium M 7x5 series processors normally run on a 400MHz FSB. However, a physical modification can be made to trick the laptop into running the FSB of such a 7x5 processor at 533MHz. The effective result is up to a 33% overclock.


Introduction:

The Processor:
There have been many iterations of the Pentium M processor to date. The original versions, known as Banias, came with 1MB of L2 cache, and was built on a 130nm process. Since then, Intel has developed the newer Pentium M, dubbed Dothan, with 2MB of L2 cache, and built on a 90nm process. At the same time, Intel has also instituted "Processor Numbers", a way for the average consumer to gauge a processor's performance. For the standard Pentium M, these can either be in the series 7x5 or 7x0.

The 7x0 processors already run at the faster 533MHz FSB, and thus this guide will not provide any additional performance for owners of 7x0 series processors. The 7x5 processors, on the other hand, run at the slower 400MHz FSB, and is what this howto will be targeted at.


The Chipset:

The purpose of the chipset, relevant to this modification, is that it controls the speed of the FSB. There are two chipsets being used in conjunction with the Pentium M processors: the 855 and the newer 915. The 855 series chipsets only support a FSB of 400MHz. The 915 chipsets can operate at either 400MHz or 533MHz, and thus is what this howto is based on.


The Processor Pin-out:

The chipset determines which FSB to use based on the value it "reads" from two pins on the CPU, the BSEL[0] and BSEL[1]. Below is a table listing which BSEL values lead to which FSBs.


Table from Pentium M Datasheet

The difference in BSEL configuration between 100MHz (400MHz quad-pumped) and 133MHz (533MHz quad-pumped) is that the BSEL[0] is L (low resistance) instead of H (high resistance) for the 133MHz FSB. Therefore, in order to increase the FSB, the BSEL[0] pin must have it's resistance lowered. To do this, it can be connected to a ground pin. Below is the full processor pin-out, in top view.


Image from Pentium M Datasheet

As seen from the pin-out, the BSEL[0] is pin C16. Directly above it, at pin B16, is a VSS pin, which is a ground pin. Thus, if the two of these pins were connected, the resistance would be lowered, and the motherboard will be tricked into running the processor at the higher FSB.


The Software:

After an overclock, stability is always a concern. To test stability, Prime95 will be used to stress test the newly overclocked CPU. If it detects an error in its calculations, the system is unstable, and needs to be downclocked slightly.

At this point, the program RMClock will be employed. Because of the way Speed Step works, all Pentium M processors are downward multiplier unlocked. This means that all multipliers below the highest one set at the factory are selectable (down to 6.0x). With RMClock, lower multipliers can be selected, in 1.0x decrements, which translates to 66MHz per decrement (0.5x133MHz=66MHz). After each decrement, the system is then retested with Prime95. This process is repeated as many times as necessary, until a stable system is obtained.

Also, this is an additional section that covers undervolting the CPU. When the CPU is made, Intel chooses what voltage for it to run at. This voltage is scalable, depending on what clock it's running at (600/800MHz or maximum speed). Intel sets a very modest buffer, giving the CPU usually far more voltage than it needs to operate in a stable manner. RMClock also allows for the manipulation of the voltage of the CPU (downwards). There are two main advantages to this. First, it would create less heat, and thus the fans don't have to run as fast. The system should generally run quieter with a decent undervolt. More importantly, with less voltage, the CPU is using less power, and therefore the battery life will be improved.


Hardware Requirements:

Pentium M Dothan 7x5 series
Intel 915 series chipset


Procedure:

The Actual Wire-mod:

To start it off, we need to open up the laptop to get to the CPU. Refer to any tech sheets for the specific laptop on how to do this. The CPU should be held in by a screw-like mechanism. Remove the CPU to expose the actual socket.

To connect the BSEL[0] pin to a VSS pin, we could wrap a wire around the actual pins on the CPU. However, once you've taken a look at how small and close the pins are, you'll find that this can be a bit of a challenge. An easier approach would be to drop a piece of wire into the two holes on the socket corresponding to these two pins. This will have the same effect, and will be significantly easier to do. The best wire to use for this is from typical electrical wire. For example, the wire for a lamp. Inside the plastic layer is a lot of thin copper wire. A short, maybe half-centimeter piece of this copper wire is all that is needed.

The top view of the pin-out is the same as a socket view. Thus, the socket needs to be oriented so that the corner with the missing pin is on the upper left.

The two pins are on the second and third row, and are just a little further than midway across the socket. The easiest way to put the pin in the socket is with tweezers. Form the copper wire into a U-shape. Estimate about where the two pins should be and drop it in. Now, count the pins across from the left (Note: the first row is missing one pin on the far left). The target is 16. Now, you should have some sort of reference as to how many spaces you need to move. Adjust the position accordingly. Once you can count 16 from the left, you may want to verify that it is indeed in the right two pins. To do this, we can count from the other end back. Counting down from the right-most pin (26), you should be at column 16 once you arrive at the connected pins.

That's about it in terms of the actual BSEL0 pin-mod. Next up is reassembly. But first, there's the option of changing out the original thermal interface material for some of better quality. There are most pro's and con's to this. A pro is definitely a cooler running CPU, since a good thermal compound would allow better contact between the CPU core and the heatsink, and thus better heat transfer. A con would be for those who are afraid of "voiding" their warranties, since changing the thermal material will be clearly noticable, should any warranty work be done. However, I doubt that that's something they would be checking for.

Now is the moment of truth.. will it boot up? Will it be stable? Let's find out. Because the Pentium M's all have a low-end multiplier of 6.0x, it originally ran 6.0x100=600MHz in its slowest mode. Now that we've increased the bus to 133MHz (533MHz QDR), the new lowest speed should be 6.0x133=800MHz. The maximum will be 133% of whatever the original rated speed was. For example, a 1.5GHz (15.0x) will now be 2.0GHz (15.0x133=2000MHz). Below are screenshots of my CPU operating at its new speeds.


CPU-Z screenshots showing overclocked minimum and maximum CPU speeds

As the Specification states, the CPU is a Pentium M 1.50GHz (model number 715). That tag will not change. However, that does not mean that the CPU is running at 1.50GHz. As long as the FSB shown is 133MHz, the wire-mod was successful. This does not mean, however, that the system is stable.

Stability Testing:
Next is the test for stability. For this, we'll be using Prime95. We will be performing a torture test (Options -> Torture Test), of in-place large FFT's. This test is set to run forever, assuming no errors arise. Ideally, it should be run over a period of 24 hours to be reasonably certain of stability. However, an overnight test should suffice for all intensive purposes. If the test is still running at the end of your period (whether it be overnight, ~8 hours, or 24 hours), you should be able to conclude that the system is stable.


Software Manipulation:
So what if your system isn't stable? Here's when we load up RMClock. First, we want to set the general configuration as follows:


Screenshots of RMClock

These settings allow RMClock to load on startup and activate any settings, and run in the background. If we click on the Management tab, we can actually set some values.

As you can see, my CPU runs at a maximum multiplier of 15.0x, which is the default max for Pentium M 715's. However, if your system failed the Prime95 torture test, the Maximal FID is what you want to be lowering until your system is stable (as per the Prime95 testing).

Make sure that you have "Run HLT command when OS is idle" checked off. Note that this will cause task manager to report CPU usage at 100% all the time. You would therefore have to rely on RMClock's Monitor tab for accurate CPU load information.

If the system is stable, we can try undervolting. This can be done by adjusting the minimal and maximal VID's. Assuming that the laptop is going to be plugged in during this testing, we will first adjust the AC Profile to Minimal. This will force the CPU to always run in Minimal mode (Minimal FID and Minimal VID). With each voltage drop, a little Prime95 testing should be done to verify stability. Once you hit a voltage that results in Prime throwing errors, increase the voltage by one increment (or two if you want a bit of buffer). Next, do the same with AC Profile set to Maximal to configure undervolting while the CPU is running at maximum speed.

The CPU I used in this test ran with the following original VID's:
Minimal VID: 0.988V
Maximal VID: 1.340V

The best I could get with maintained stability is as follows (as seen in the picture above):
Minimal VID: 0.700V
Maximal VID: 1.148V

This information is just provided for comparison purposes. Your individual results may vary, of course.


References:

Prime95
RMClock
CPU-Z
Pentium M Datasheet
Back to top
View user's profile Send private message
ziddey
Site Admin


Joined: 05 Mar 2005
Posts: 3

PostPosted: Mon May 09, 2005 3:14 pm Post subject: Reply with quote
Update:

A new program, Centrino Hardware Control (CHC), can be used to replace RMClock and Prime95. CHC is able to undervolt and control the maximum multiplier, both of the features RMClock was previously used for. It features two levels of stability testing (quick and full tests), which can be used instead of Prime95. Furthermore, CHC comes with some nice thermal and other monitoring tools, which come in handy. Also, unlike RMClock, which causes the CPU load to be reported as 100% all the time, the CPU load is reported correctly with CHC.


CPU Voltage tab of CHC

CHC is available at http://www.pbus-167.com/chc.htm


Update:

I've done some testing with CHC. I have my configuration set so that it's dynamic switching when on battery and max performance when plugged in. With CHC, it seems to have some difficulty going back to full speed after I unplug and replug my notebook. As of such, I've resorted back to RMClock for now.

*/