Removing the IHS from the Northwood-D

Howto by Liquid3D @ 2003-09-09

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.
  • next


Warning the pictures in this HOWTO are for a Mature Audience, no one under 18 admitted, without a volt-meter!

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.
  • next
Comment from FreeStyler @ 2003/09/09
If you can get some of them foamy pads AMD uses, they sure help. The core is not as brittle as AMD, but it's not invincible.
Comment from some1x @ 2003/09/09
I removed the IHS accidentally :-0.

What happened is that the Arctic Ceramique caused the heatsink (SLK800U) to stick to the IHS. When I removed the heatsink, the IHS simply came off with the heatsink.

The biggest concern I have is that the IHS added thickness to the CPU; and the heatsinks are designed with that thickness in mind. So now, the heatsink applies much less pressure to the core. (I wish I knew the spring constant for the SLK800U's springs, then it'd be possible to calculate the difference in pressure).
Comment from TeuS @ 2003/09/09
some guy removed the IHS, lapped everything well and placed the IHS back (with some good thermal paste offcourse).

without the IHS the CPU ran 1° colder, but when he placed the IHS back his temps decreased 5°. quite impressive
Comment from Unregistered @ 2003/09/09
What'd you use to remove the epoxy or whatever it was holding the IHS on?
Comment from Unregistered @ 2003/09/09
pure 100% alcohol
Comment from jmke @ 2003/09/09
an Alternative method using a knife:

As for results, I've been running my machine now with my IHS-less processor for over three hours straight now, all the while running Prime95. My processor is currently sitting a full 2°c cooler than before I removed its IHS
Comment from Unregistered @ 2003/09/09
This sure is awesome. I have been sitting around for a year learning from others here and there. I have a 2266 B0 stepping and it goes 3060 with air. I am within weeks going to watercool this one and I think I wanna remove my IHS aswell.

I have a friend @ who can make me a special head. However I need to know how thick this heatspreader is, anyone?

Kind regards Martin
Comment from biCker @ 2003/09/09
This should also work for a 2.26 533fsb, right??
Comment from jmke @ 2003/09/09
Yap, without problems, but start by just lapping the IHS, Liquid3D had already some good results by only doing that!

Prooof is in the sanding. Yesterday in the same Ambient temp (my air conditoned room) my overclocking temp was 45C to 49C at 3.6GHz (300FSB) today same ambeint temp, and OC conditions, and the temp's are now 38C, and haven't gone above 39C. That's 6C to 11C temp drop!
Comment from biCker @ 2003/09/09
right, but his is a copper one, I don't think this is the case on 533fsb cpu's. However, it is worth the shot.
Comment from Unregistered @ 2003/09/10
Comment from Unregistered @ 2003/09/10
Um... all Intel heat spreaders are copper just like how they all had hyperthreading capabilities built in but was hardware burnt and disabled permanantly.

My question is why would this decrease the processors life by 50-75% like he said?
Comment from Liquid3D @ 2003/09/11
Thank you. I was in error printing that. 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. A greater chance the core would be cracked by a heavy heatsink or mishandling. I know of no reason why simply removing the IHS would have any detrimental, electrical, or other negative effect. In fact if the chip is handled carefully then common sense says that any few centigrade reduction in overall operating temps would actually lenghten the life of the chip, and by a significant amount.

In so far as the copper question. I think their (IHS's) all copper underneath. It's misleading but since copper is third best conductor of heat, I'm sure Intel has been using copper for some time. There's really no other metal; except aluminum, due to cost.

Here's an example;
Thermal Conductivity is the amount of heat a particular substance can carry through it in unit time. Usually expressed in W/[mºK], the units represent how many Watts of heat can be conducted through a one meter thickness of said material with a one degree Kelvin temperature difference between the two ends.

Diamond = 2000 - 4000
Silver = 417.3
Copper = 393.7
Gold = 291.3
Aluminum = 216.5
Steel/Iron = 66.9
Lead = 34.3
Ice (H20 @ -0.5C) = 1.6
Carbon = 1.6
Glass = 0.8
Comment from jmke @ 2003/09/11
Thanks for the clarification !
the howto has been updated,
Comment from Liquid3D @ 2003/09/11
Originally posted by some1x
I removed the IHS accidentally :-0....The biggest concern I have is that the IHS added thickness to the CPU; and the heatsinks are designed with that thickness in mind. So now, the heatsink applies much less pressure to the core...
I'm glad you brought this up. I had the same question even after lapping the CPU. What I did was to foolishly lap down the standoffs, and the screw bottoms (they tap-out in the standoffs) to try and compensate for the height reduction. All I really did was ruin my SLK-900-U. Luckily I had the SLk-947-U in for an article, and was saved. I'm donating the SLK900U to a freind in canda, who's going to try mount a fan on it's side pointing down? I still haven't figured out what he's up to. It was also brought to my attention, simply placing a few washers under the springs would have the same effect. I haven't figured that one out yet either but read on.

Here's the pretty part. I then realized the reduced height was just enough to reduce the pressure, thereby avoiding "my" biggest fear, cracking the core. My temps have been at 29C for the last four days running my 2.4C@3.0GHz on 250FSb 1:1, at 1.550Vcore. Their currently at 28C as evening falls. I beleive the reduced height is the perfect ad hoc blessing which solves the problem of possibly damaging the core. My answer; do nothing. Mount your HS carefully, and IF the temps seem high, then you can lap the bottom of the screws, and the standoffs they sit in, because they tap-out in the bottom of the standoffs, so I lapped both.
Comment from Unregistered @ 2003/09/11
I'm surprised that you think a "4-6c" drop in temp is insignifigant -- article is worded that way... Everyone's switched to more expensive PCM+ for a 2c difference on the average....another 4+c difference for a little work is big dividends in that context.

Exposed core protection; anyone check the thickness of a shim for the Radeon cores? I don't have a good pair of digital micrometers, but the one from my 9800Pro looks awefully close...
Comment from Liquid3D @ 2003/09/11
I guess your right about that, 5C-6C is very significant. I don't know why I was minimizing that? And as a matter of fact, if done well, temps will be even lower. Based upon my tests since removing the IHS, temps have dropped as much as 8C! I'm currently using AS3 until I determine what's happining with Nanotherm PCM+, (this is what I'm refferring to; I didn't want to use PCM+ just yet. But if PCM+ works as well as it has, then my current temp of 28.50C will drop further to 23C!!!

Here's a shot of my current temps at default speeds (yet 1.50Vcore).
Comment from inteks @ 2003/09/12
i did it too.
here my results.

p 2.4B 533fsb

18*156 @ 1.675V (prime95 aproved ! 12houres)
idle around 53°C stressed till 72°C

18*160 @ 1650V (prime95 aproved ! 12houres ,too)
idle 45°C strressed 62°C

maybe because the .025V voltagedrop but i dont need more vcore now!!!

i did the method with the cutterblade 10min of work.

but i has to remove the mainboard because i have a alphacooler.
after removing ths IHS i have to remove the plastic disks to get the cooler contact the cpu.

have fun
Comment from inteks @ 2003/09/12
maybe i can go a bit higher then 160 now but i like flat numers !!!
) 2.88 GHz flat is ok!!!

here is a pic:

Comment from Liquid3D @ 2003/09/12
WOW Those are excllent results! Very good. Of course now it's time to choose what paste will be ideal for your newly exposed core?

As a result of my Thermal paste article, I've recieved many educational emails. One thing I think I need to clarify. I made some comments inferring Arctic Silver may have been "laxidaisical" in their R&D. Someone who happens to know a great deal about the industry, and even about Arctic Silver made a very good points. Arctic Silver has not been reticent in their research, in fact ; "...they do more R&D then Nanotherm will ever do, or even be capable of..." He made a very astute point there, he also said, "Arctic Silver can't afford to release a product until extensive testing is done based upon their customer base. While Nanotherm is a relatively small company, and not nearly as large in their distribution, AS is much larger, not to mention the clientelle which Arctic Silver is a supplier to (i.e. NASA). Arctic Silver's paste must hold up under the most demanding conditions in the industry..." he also mentioned PCM+, and said I should take any old Aluminum heatsink I have, apply a few drops of PCM+ and put it in the oven at 350F. The take some Arctic Silver 3, and do the same, and I'd see some interesting results. I've not tried this yet. So without further adeau here's ome temps with PCM+ on my 2.8C at 250FSB 1:1, and 1.575Vcore (I found the 2.8C needs a litlle more Vcore to remain stable);
Comment from inteks @ 2003/09/12
i use AS3 and its ok.
the temp's are lower that i ever think to get to.

i have done an extremtest (befor i remove the IHS)
i hold the fan and run and ut2003 bench.
the cpu slows down @ around 90°C. after i release the fan the cpuspeed goes up again.

you find many threads about the problem that p4-mainboard doesn't show's the correct temp.

i think the slowdown occur at 72°C that's the temp in intels datasheet

so if my mainboard reports temp's 20°C to high. then its 25°C for real (45-20) in idle!!! and i think i cant go below casetemp. thats impossible.

Today 16:32
btw: 166 is also a nize number!!! but prime95 fails at this speed!!!
i can play ut2003 till 173Mhz FSB

have fun
Comment from jmke @ 2003/09/12
eXtreme indeed, talk about a cripsy burn-in at 90°C
Comment from inteks @ 2003/09/12
Originally posted by jmke
ps: eXtreme indeed, talk about a cripsy burn-in at 90°C
i just wanted to know how far the cpu goes.
normaly no cpu runs stable at 80°C !!!
and the cooler was handwarm !
Comment from jmke @ 2003/09/12
Comment from BigBen2k @ 2003/10/29
I have a favor to ask...

If anyone has saved their 'removed' IHS, either from an Opteron or from an Intel processor, please PM or e-mail me the dimensions, will ya? (thickness is the figure I'm after!)

Alternatively, I'll cover the shipping (PayPal?), if you can send it to me!

Thanks to all! (Nice to be back )
Comment from Liquid3D @ 2003/11/10
Sorry about the delay here's the pic you asked for. The edge of the IHS is twice as thick as it's entire surface. So if the edges are 2mm thick, the IHS is 1mm in thickness. Let me know if you need more photo's. I'm thinking about sedning this one to you, but first I must decide if I'll be removing another.

I want to put AS5 on the core and replace the IHS using AS3 Epoxy kit. I surmize this is the reason your looking for a replacement? You removed it, tossed it away, and now your having problems with heatsinks having stand-off's as the difference in height now prevents contact between P4 core, and heatsink?

You can sand down the stand off's 1.2mm, and you must also take that amount off the bottom of each screw as they "tap-out" in the bottom of the stand-off (nut);
Comment from Unregistered @ 2004/01/31
Liquid3D......U hav an email regarding the P4 Heat Spreader.
Comment from Liquid3D @ 2004/02/01
Got it. Here's an idea of the height of the P4 with/without IHS. For the SP-94 to seat properly, approximately 1-mm will have to be removed from the top of each standoff. Because the sprimg-mounted screws "tap-out" in the bottom of the stand-off's (which was prevents people from overtightening, hence damaging the chip) they will have to have approximately 1-mm sanded off from their bottom;
Comment from Unregistered @ 2004/02/01
Liquid3D....the term regarding "stand-off" refers to the 4 screws or the mount bracket below the Mobo ? And we have to grind / sand 1mm off as well ? A pic refering to wat we have to alter would be the best. Thanx and sorry for the trouble.
Comment from Unregistered @ 2004/02/01
Disregard my last post bro....the email you sent me is very detailed. Thanx Alot.
Comment from Liquid3D @ 2004/02/01
"Stand-off's" refer to the four receper bolt's which themselves thread into the black-plate of Thermalright SLK and SP-94 type heatsinks. These stand-offs which screw into that mounting X plate, accept the mounting screws which hold down the heatsink. You have to remove about 1mm from each, and the from the bottom of the screws which screw into them. This is to lower the total height of the heatsink to accomodate the altered height of any P4 with it's IHS removed. Here's pic's of the stand-off's and their corrosponding screws;
Comment from Unregistered @ 2004/04/15
what ever happened to the 2.4c with the removed heatspreader? any results on that?
Comment from jort @ 2004/06/09
i removed also the IHS of my p4 2.66ghz
its looks very cool
here are some pics.bad quality pics:-s
Comment from jort @ 2004/06/09

Comment from Liquid3D @ 2004/06/11
Here's a close-up of the 3.0C core, and my 2.4C with IHS removed gave me another 30Mhz, and about 5c overall temp drop. The 3.0C did even better with a 200MHz OC differential, and about 7C drop in temp. I imagine this is due to the convex, or concave imperfections in these IHS.

No matter how their applied (by machine) they still vary due to micro-imperfections in metal, and TIM thickness. If just a few more milligrams of TIM is placed between IHS and core, this could lift the IHS just a few tenths of a milimeter higher, or lower. Point being it's not a perfect manufacturing system by any means and the fact it's taking place in plants thousands of miles away doesn't help the consistancy factor. The fact some Malay's overclock better then some Costa Rica's indicates this. My newest project is to consider removing both the A64, and Prescott IHS's.
Comment from Sidney @ 2004/06/11
Man, it takes the eyes of a spider to clear up the fuzzy pictures. Is the ground moving or is it me? :grr:
Comment from Liquid3D @ 2004/06/11
That's why I published my pic's. I thought I'd assist "Spidy" in clearing things up. After all he's so often misunderstood.

On the right is my 3.0C IHS removed, on the left the 2.4C. I forgot to clean them up, they were in hybernation. So pardon the dust, huhuhuhemmm......


By the way Jort, although your pictures are blurry, you did an excllent job of removing the IHS's! Look's like you cleaned them up nice. No traces of epoxy like on mine. I had to leave extra on one though, so I could tell them apart...
Comment from jmke @ 2004/06/11
Originally posted by Liquid3D
so I could tell them apart...
use a black marker to write their specs on the top
Comment from Liquid3D @ 2004/06/12
Doh...good idea
Comment from jort @ 2004/06/17
tnx for the possitive coment i did some time to remove it
i did i with a friend and some "garendraad"