I believe the method described in this howto, may be the safest (albeit time consuming) for removing the IHS (Integrated Heat Spreader), from your P4.
In our endeavors as Overclockers we have a propensity for continually pushing hardware to its limits. When voltages and BIOS settings no longer suffice, we begin physically modifying our hardware. This led me to tearing down my P4 2.4C Malay in stages, if you will. Initially I lapped the IHS (Integrated Heat Spreader) to its copper, and then went back and did it again.
Finally I decided to go all out. This is a learning experience for me, and although it's difficult for me to obtain hardware for testing, I believe it's our curiosity which is at the heart of PC-enthusiasm. And while Intel employees may be pulling their hair out, this article in no way condones overclocking, or modding. Obviously either will violate your warranty. This is purely an educational experience. Because there are people who will do it, regardless of risk, I thought I'd try and give the process some integrity. However, truth be known, I do not believe it's necessary for anyone to go to these extremes, which is why I've gone through them for you. Intel's IHS performs its task very well, adhering perfectly to its intended design purpose.
Most people who have performed this procedure, have only gained 4C to 8C drops in temps, while there's a 50% - 75% reduction in your chip's life afterwards. The odds are clearly unbalanced, and the logical person should avoid this.
Update, clarification of the chip's life reduction:
This was solely based on a mechanical mishap, like cracking the core under the pressure of the heatsink. What I should have said is there's greater chance say 50% if handled improperly the chip could be cracked at the core, through mishandling. Or under the pressure of a heavy heatsink.
I didn't mean to imply removing the IHS in itself would have some other ramification other then the chip being exposed to the elements. I know of no reason why simply removing the IHS would have any detrimental, electrical, or other effect. In fact if the chip is handled carefully then common sense says that any few centigrade reductions in overall operating temps would actually lengthen the life of the chip, and by a significant amount.
Given the recent overclocking success of the Northwood-D, Intel's manufacturing quality has been exceptional. I've yet to find in my entire overclocking endeavors a processor with the innate performance potential of the Northwood-D, and particularly the Malay 2.4C's. And this design predates SSOI, which is currently being implemented at the .09-micron die, in the Prescott series. I've researched SSOI, and its potential for ameliorating static current leakage, hence reducing thermal output, will be significant. And while Silicon On Insulator technology is an excellent thermal management solution for .09-micron, it offers many more attributes.
Until someone builds a better mouse-trap, current microprocessor design is based on millions of CMOS-type switches, or transistors. These are capacitance switches; in order to function they must be constantly charged. What is inevitable is some small amount of current will discharge into the surrounding silicon because of its semi conductive properties. If one can insulate those areas at the switching junction, where junction capacitance occurs, one would eliminate leakage. This is where the term Silicon On Insulator derives from. In placing a Silicon-oxide, or glass insulator between the treated-charged silicon and the silicon substrate, this phenomenon will be eliminated, or at least reduced. This will speed up the transistor, and of course reduce heat. Since there is always current present, there will always be heat associated with it, but any reduction, especially one as significant associated with SSOI, is welcome.
Intel's implementation of SSOI should have a great impact on their Prescott beyond the prima facie speed increase associated with the die shrink. Obviously if transistors are closer together, and more can be packed in smaller areas the processor will not only be faster, but have more performance potential. There may even come a day, when overclocking isn't necessary (as many already believe). After all, were going to see 5GHz in the next year, so I don't see how much more overclocking will accomplish. Truth be known, were not really overclocking anything!
Based on the manufacturing process, the first model's in any CPU line, are from the same core design and architecture, which will eventually power the fastest in that line. Therefore, all Overclockers are really doing, is finding the CPU's performance ceiling prior to the release of the fastest model. And beyond that, their simply exploiting what engineers have already built in. We’re never going to get much faster then what Intel's Q&C methods will produce. After all it's their (Q&C test engineers) sole purpose to push the chips to the absolute maximum, to the point of failure. In knowing these limits they can accurately determine at which speed their fastest model should run, while still giving the consumer years of dependable, reliable performance.
Of course there are those of us who circumvent certain aspects of the design, and employ such materials as Liquid Nitrogen to super-cool the chip. Then pump as much voltage as the motherboard (after modding) will allow. These are the true Overclockers, because they probably do find speeds even the Intel technicians do not. However, not even I can tell you if there are containers of LN2 at Intel's Q&C facilities. I'm willing to bet there is. Some of the more successful designs are built on extremely low operating temps. One such pioneer is from Denmark. Chip-Con's Prometeia line of Phase-change coolers, which drop the constant operating temps below freezing, are perhaps the most effective yet.
They are reliable, dependable, and currently in the hands of the more astute overclockers. I have one myself, although it needs to be RMA'd.
So without further a due and I hope the above has been educational, let's get into exposing my Malay 2.4C L310A735-0101 (OEM) for the entire world to see.