Lapping the i875 NB

[Liquid3D]

I decided to try lapping the NB on my Epox. I foolishly forgot to take photo's and I've put the dang passive HS back on there with the last drops of AS3. Regardless I can stiil take some as I haven't put the Epox back into my system yet. I have to run some memory tests on this IC7-G again. Onto the subject at hand.

I belive it was MajorSlaughter who suggested; ...lap the NB-HSF on the IC7-G, and he was abosolutely correct! Originally I couldn't do it because I didnt' have the paper, but I just did. It took almost an hour just to remove the residual glue left behind from the "foam-like" shim, placed to protect the NB-chip (which I feel only inhibits any airflow). The heatsink bottom is very coarse. The residual glue just fills what are surprisingly deep groves left from a rough sand job. I can see what he meant by the poor contact, which results in poor thermal transfer, raising temps, (and lowering memory performnace). There must be air-pockets trapped in there. And the pink TIM manufactures are using, seems to have a consistency of rubber. When a TIM has this much elasticity, it's integrity as a true TIM, is in question. I beleive there's a two-fold purpose manufacture's use this. This is about the tenth time I've removed this stuff from NB/HSF's, and it must double as a bonding material, which would exaplin it's rubber-like quality.

I then began to alp the NB, and using 600-grit, was ever so gentle. I removed the blue coating, onpon which the chip's ID is printed, until it was silver. The reason I posted this was to seek adivce from those more experienced with this. I've damaged NB chips before by accident, and know how easy it is to expose and damage trace wires. When attempting to remove the stubborn pink-TIM once I foolishly used a flat-head screwdriver. I thought i was being careful, but exposed the trace copper interconnect wires. Big mistake.

I havent put the Epox back in to see what differences there will be as I have to test this IC7-G, right now. I didn't lap the NB on this board because at this point it's technically not mine. But lapping the NB-HSF, obviosuly wil not harm the NB, but in fact increase the thermal transfer. I used the tiny amount of AS3 i had left. So I'm about to run some benchies. I'll have to take the Epox out of the box, and take some Pic's for ya. But as I said at this point, I removed the blue, and am down to aliminum. Can anyone tell me how far I can go safely? Is the NB core covered with a copper or aluminum casing?

[Major]

It doesn't have any copper in it so don't try to get down to it.

I only lap it enough to get the blue coating off. I wouldn't do too much more than that, the die is near perfetly flat anyway, but IMO removing the coating would help in thermal conductance.

[Liquid3D]

Thanx Major. You know how I've been lapping my IHS. Well I wanted to get it even nicer so i did it a few more times, and bit more on the HS. Well when i was doing it I thought it might have adverse affects with stand-off's on SLK900U. Someone else mentioned this as well. SO I lap her down, and apply the correct amount of PCM+ (Nanotherm say just two drops will do). Put it all back in and my temps were 39C!!! When I first ran this chip in my IS7-E I was running at 33C!!! And that was at 3/6GHz.

So I remove, and really lap down the stand-off, and of course the seating-screws, which also must be sanded or it's useless. So I do this propbably as much as I'd lapped the HS, and IHS, and now my temps are freaken 107F/42C? Ya know what could be up? I put a little thicker coating of PCM+ this time. It seemed to work best when I put about 6-drops on there, which Nanotherm said was way too much?

t ya think Major? imagine if i removed the IHS completely? I got a peice of strong thread to do it with as well. Apparantly the safeest method is to "saw" through the epox with a piece of thread.

[Major]

I dunno about the temps, I took the IHS off my 2.8 today and I had a hard time getting the vapo cooling head to make good contact. The temps where way up at first. I am able to sqeeze just a couple of more fsb out of it now so it did help some.

It would have to be a contact issue, either not making 100% contact between hs and core or not enough contact pressure. Also I ended using the lightest application of Cerimique that I had ever used, and my normal aplication is pretty sparise. Just enough to dull the shine if you know what I mean, didn't even appear white.

Edit: to remove the IHS I just used a razor blade and worked it in between the IHS and chip substrate at each corner, sorta rocking it back and forth (not up and down but sideways where it was cutting the glue) and after a while of doing that one corner came loose and then another and at that point you can just pull it off. Its alot easier than you think, I was sweating bullets !!!! lol

[Nanotherm]

Liquid3D,

Thinner is always better, but if 6 drops works better for you - it works for me