Hi everyone!

Just thought I�d put a link to my brief guide to using these processors in desktops motherboards. I have to admit the main reason I did it was to offer a guide for people buying them from me off eBay (I've posted a link to my auction on my site and also in another post over in the Deals, Vendors & Classifieds forum so as not to break any rules or anything).

Anyway I have to thank you guys for sharing your wealth of knowledge. As soon as I get our little cube PCs for work I'll let you know how I get on using these CPU's in them.

Cheers,
Will

Good overview. Gets a place in my link list

About your Vcore problem: When comparing 1,25V and 0,8V at same speed there should not be any temperature difference. Correct me if I am wrong.

Thanks for reading it, all comments (especially positive ones) most welcome I know it doesn't really compare to the very thorough SPCR reviews/guides but if I get chance to make it bigger and better I'm sure it will become a useful guide for many.

I would expect there to be a huge difference given that many people on here rate voltage as much more significant on temperature than clock speed. Also, as mentioned I would be very surprised if I could get up to 2.26 on such a low voltage. More testing is order but any help or tips as to some thorough testing method would be most appreciated. I am kind of limited with equipment but I'll try my best.

The guide looks good! Quite impressive.


Yes, changing vcore, even at the same speed, makes a huge difference in wattage. Wattage increases at the rate of the vcore increase squared. (double the vcore=4 times the wattage) Speed and wattage are directly proprtional, double the speed and the wattage only doubles.

To demonstrate from your CPU temp test:

1.2Ghz @ 0.8375v=13.3watts
1.6Ghz @ 0.8375v=17.7watts
2.26Ghz @ 0.8375v=25.0watts
2.4Ghz @ 1.3375v=67.8watts
3.2Ghz @ 1.68v=142.7watts ( )

Actually, after looking at your numbers, I think your suspicions about the motherboard are correct: it is only running as low 1.25volts. The math at 0.8375 doesn't work, given the temps. (the full explanation, with the math, is a bit long, but if you'd like to see it, let me know, I'd be happy to type it out)

Can you pin mod the P4-m to reduce the voltage beyond what the BIOS will allow?

Nice work!

Thanks for the kind words!

I guess setting the voltage through pin mods would be one way of doing it although that would make it permanent which I would prefer to avoid. There is also the possibility that the GA-8I875 just doesn't support a VCore as low as that so even with a modded CPU it might still supply 1.25 or even worse, the default 1.5!

Mind you, I have a 1.6GHz Mobile P4M which I broke 3 Vss pins off by accident (don't ask) and yet it still works. This may the ideal candidate for a bit of testing!

My point was: If you mb regulation is broken, there should not be any difference.

So you're manipulating speed by adjusting the FSB; and adjusting Vcore to suit (for stability, reduced heat, etc.).

Is it possible to adjust the multiplier? If you're settling for a lower speed (for heat reasons) wouldn't it be better to increase FSB as much as possible while reducing the multiplier, as opposed to simply decreasing FSB? (And does this question even make sense? It could be completely naive -- I have no experience with these kinds of adjustments).

--
Aidan

It's not possible to adjust the multiplier on Intel CPUs. They are locked from the factory, except for a few, difficult to obtain engineering samples.

Cheers for replying Ralf.

It would be great if you could maximise the FSB potential while adjusting the bus speed multiplier to suit thermal requirements. However, I think the x12 multiplier on a 200 MHz FSB is quite a good compromise between performance and heat.

Perhaps a point to note if we could push the FSB really high and reduce the multiplier is that considerations have to be made for (silently) cooling the north bridge the further you push the FSB. As I’m just testing on the bench at the moment I’ve left the fan on my GA-8I875 to keep things cool. I’m thinking of replacing it with a Thermalright NB-1.

Thanks for your comment, Ralf. I have picked up a lot of knowledge from your wonderful posts to this forum.

Another possibly naive question: I remember reading/ hearing that laptops adjust the speed of P4-mobile chips on the fly, by changing the multiplier setting. I wonder why this doesn't translate to an unlocked multiplier when the chip is placed in a desktop motherboard (as is the case with the Athlon mobile chips)??

I'm trying to clarify what the advantages of a mobile P4 chip would be for silent computing purposes: If I put a mobile P4 into a desktop board, and it defaults to a multiplier of 12, and I run it on a FSB of 133, for 1.6Ghz, I assume this will be running cooler (and a bit more efficiently, with a bigger FSB) than an equivalent desktop P4 (at around 1.6-1.7 Ghz)? So it might be a nice way of getting a slow to moderate paced P4 at a cooler temperature? Would that be pretty much the main potential benefit? (I understand it's also possible to over-clock, although I guess you'd be moving out of 'silent' territory then, with greater heat output).

--
Aidan

That little chipset cooler is 'adorable' -- just like a miniature version of it's bigger brothers.

--
Aidan

Sorry to dig up this old thread.

This looks interesting.

Will this work on an asus p4p800?

kogi

Certainly.

A mobile pentium 4 should work in that mobo.

A word of warning about the Asus p4p800 or any Asus for that matter. They normally lock all their bios from running vcore lower than the default 1.5v.

I notice you're using a GA-8IPE1000 Pro with your P4-M chip. How are you finding that motherboard with a mobile P4? Any advantages or disadvantages worth mentioning?

The GA-8IPE1000 Pro has a feature which Gigabyte calls CIA -- Cpu Intelligent Accelerator. This is an overclocking feature, which "turns on" when the CPU is at max. use. This seems like an ideal feature for a P4-mobile chip. You could run your chip at the default 1.2 Ghz (low heat) and get a bit of a performance boost when your CPU really needs it via the automatic FSB overclock. (It's almost like a reverse SpeedStep, which drops the multiplier down when the CPU is not in use). Unfortunately the maximum overclock doesn't seem all that high (+7%) which could limit the usefulness of this feature with a mobile chip.

http://tw.giga-byte.com/Motherboard/Fil ... ch_CIA.htm
[Gigabytes explanatory page on "CIA" with a few bios pictures]

There also appears to be a way to underclock, when the CPU hits maximum use: you can choose a CIA setting of down to (minus)20%. So presumably when CPU use is at a peak, the FSB is dropped by 20% -- which might cool things down quite a bit. I wonder if this features would be of much use for silent computing purposes (preventing dramatic overheating)? Or just incredibly irritating

The gigabyte is a great board, but I did have to cut the bsel1 pin to get a 800fsb overclock. Its at the p4-m default 1.2v vcore (lower crashes sometimes).

To cut the bsel1 pin read the guide above or read an old discussion here

One other problem ive noticed and I think this affects other motherboards even AMD ones, I turned the computer on once without the 4-pin 12v connector, and from then on I get coil buzz from the motherboard. Ive covered all the coils in silicone it made no difference. The coil noise is related to CPU usage.

The motherboard does come with a fan on the northbridge but I used a zalman northbridge cooler for that.

I have the CIA and the MIA disabled in the cmos.

Sorry, I've been away and missed all the action on this thread.



Just to go back to Aidan's questions, you're correct that with SpeedStep technology a laptop can dynmically alter the CPU's host clock multiplier on the fly. Unfortunately no one has implemented a solution to take advantage of this in a desktop motherboard, I imagine because SpeedStep is implemented through the chipset.



You're pretty much right on with the advantages of running the Mobile P4-M in a desktop. Not only do you gain performance from increasing the FSB but I'm pretty sure that a Mobile P4-M will run cooler clocked any frequency between 1.2GHz and 2.4GHz than the equivalent desktop P4 running at stock speed. By that I mean if you take a 1.7GHz Mobile P4-M and overclock it to 2.4GHz, I bet it would run cooler than a standard desktop 2.4GHz P4. Perhaps a D1 core will run cooler than a C1 core but comparing like with like I reckon the Mobile P4-M would win every time. So you can have performance and not move out of "silent" territory. I really need more desktop P4s to test this out. Perhaps once Ralf gets hold of a Mobile P4-M he can test this.

As for the Asus mobo, I would stick to ones that offer undervolting.

Will

P.S. Sorry for tha spllin mstaks, no checka nd gprs costs me £7.50 / mb!

Im interested to know if a P4M would work in my desktop replacement Acer 427LC? It currently has a desktop P4 chip in but the chipset is the 845MP (mobile chipset). It would definatly make the thing run more quietly and give longer battery life if a mobile chip would work at its default voltages.

The other question I had, was how "paranoid" were you about crushing the chip's core (given that it doesn't have the protective metal thingy/ heat spreader)? Do you think it is fairly robust? Would most Socket 478 heatsinks be appropriate, do you think?

--
Aidan

Here's an interesting article on desktop overclocking with the Pentium 4 - M:

http://www.vr-zone.com/guides/Intel/P4M/

This guy has removed a heat spreader from an old Williamette P4, and is using it to protect his P4-M (along with some sticky tape, and some thermal paste). Quite inventive.


Also, here is a useful summary of the nomenclature of Intel's Pentium $ desktop and mobile chips:
http://support.intel.com/support/proces ... 007500.htm

Summary:

Mobile Intel Pentium 4 supporting hyperthreading - Have a 533 FSB. Socket 478. Range in speed up to 3.2 Ghz. These appear to have a thermal guideline similar to the desktop (standard) pentium 4. More recent chips have 1 Mb L2 cache. According to the intel spec finder, these seem to run at 1.3 or 1.5 volts (it varies), so maybe some of them are pumping out marginally less heat than the desktop variants.

Mobile Intel Pentium 4 = Have a 533 FSB. Socket 478. Range in speed from 2.4 to 3 Ghz. These appear to have a thermal guideline similar to the desktop (standard) pentium 4. These seem to run at 1.5 volts.

**Mobile Intel Pentium 4 - M = the chips we are talking about here. Run at about 30-35 Watts. Supposedly are socket 479, but fit into a socket 478. (What's the difference between the two?) Have a 400 Mhz FSB, and run stock at 1.3 Volts. Range in speed from 1.4 to 2.5 Ghz.

[Pentium 4] = standard desktop chip. Run at 1.5 volts.

IT'S VERY CONFUSING.

But at any rate, it seems like it's only the "Mobile Intel Pentium 4 - M" that would be running significantly cooler than a standard desktop Pentium 4. Does that seem right?? Or are these P4-M mobile chips pretty much the same as the other mobile P4 chips in terms of heat output, i.e., are the apparent dramatically higher higher thermal guidelines on the other mobile P4 chips (>60 Watts, compared to 30-35 Watts for the P4-M chips) are an artifact of the way the specifications are calculated?

--
Aidan

Fascinating thread on over-clocking with P4-mobiles (and also mobile celerons):
http://www.abxzone.com/forums/showthrea ... ge=1&pp=15

Achieving 3.5+ Ghz from a 1.8 Ghz P4-M chip, using air cooling, is apparently easily within reach. At slower speeds, very cool temps, even passive cooling, are possible. Very interesting chips -- I'm surprised the P4-Ms haven't attracted more attention from silent computing enthusiasts. They certainly seem to run substantially cooler than desktop Northwood P4 variants. Note that the mobile (P4) celerons are faster than desktop celerons; the mobile celerons have 256 Kb L2 cache, the desktop variants are limited (crippled) to 128 Kb.

I am using a Zalman 6500 on mine. I got the one which screws onto the motherboard (requires removal of the p4 retention bracket) -- this eliminated the need for a CPU shim. Also there is then heatsink->core direct contact.

My p4 1.4M will run 2.4ghz at 1.5V. It will run 2.0ghz at 1.3V. Its been up well over 100 days without reboot. FWIW, It runs 31oC idle (I use it on an Aopen AX4GE Max motherboard). Not sure how accurate these readings are. The fan is the 92mm zalman one running at 5V.

Perhaps it's time to do some more testing and publish the results. I have a 2.2GHz Mobile Pentium 4M with D1 stepping that is dying to be pushed to the limits Anyone care to take a guess as to how fast I can push it? I wouldn't bet against the 4GHz mark.



You and me both! Well, someone brought it to my attention in this forum so now we all know and more people can take advantage of it. Someone even contacted me for using these processors in budget PCs just because they're cheaper than getting hold of even the slower desktop P4's.

One thing I've found is that the standard heat sink retention mechanism doesn't work well with these chips. I tried my mobile chip in an Asus Pundit, and the heatsink didn't even make contact with the core of the chip. There was a gap where the 'heat spreader' usually is. Heatsinks which require removing the standard retention mechanism (and then bolting the heatsink to the motherboard) give better contact with the chip core. (Example, the thermalright products).

I have say I'm really dissapointed to hear that. Sounds like I should modify my listings to warn people about that.

Maybe the problem was with the Asus Pundit design. (The Pundit is a little barebone computer with a non-standard shaped motherboard). Perhaps try it out with a standard intel heatsink on a 'standard' motherboard? Maybe I'm jumping to conclusions.

No, the problem's with the lack of the thickness that's normally provided by the P4 IHS.

The height on every S478 socket assembly is the same from board to board and manufacturer to manufacturer so that any P4 S478 heatsink will work on any S478 board. The problem exists because the P4-M CPUs do not have the IHS on them so the overall height of the CPU package is less than a regular P4 (non-M) package. The standard heatsinks don't have enough clamping force to contact the IHS-less CPU die so the heat isn't transferring correctly from the CPU to the heatsink.

You'd have to use a heatsink (like the Scythe Samurai or Thermalright) that has no downward limitation to it's travel, or use the corrrect thickness of shims under the clips on a heatsink like the Zalman 7000, or use a "bolt-on the board" style of heatsink like the Alpha and mill off the correct amount from each of the four brass standoffs.

I wonder if the shims you can get for AMD chips would be useful for this?

Another solution I read about was to use an integrated heat spreader pulled off a dead P4 chip. This would add a bit of protection to the core, but it would also, I'm guessing, not transfer heat quite as effectively.

It would be great to make the most of that direct contact with the core, so obviously copper would be ideal here. Can anyone suggest particular heatsinks which offer unlimited travel as Ralf describes? I had my eye on a Spire SP420B8 Copper Stream until I was alerted to this problem; now I don't know if it's suitable.