Lapped core 2 duo IHS pics

[celemine1Gig]

Well, it's not a question of being "as xtreme as possible", most of all because I never lapped a CPU in my life, nor I have the time or extra money to do it.

Don't worry mate. You didn't miss more than aching hands covered with metal dust stains. And all that for a few degrees and perhaps some more MHz, if even at all.
I have to admit though that I still do it regardless if it really helps or not. It's just that I somehow have to do it.

I was just remembering my Physics classes. I mean, if you don't have holes and crevices for the TIM to fill (which is why TIM exists in the first place...) when you lap a CPU and heatsink, and this doesn't seem all that farfetched, according to what I've read here, then one of two things happens:

1) The TIM has a lower heat conductivity than copper, meaning using it will end up insulating the CPU-to-heatsink heat transfer, actually being worse using it than not (TIM is used to fill holes, providing contact points where they don't exist, meaning they will ALLWAYS have better heat conductivity than air; if there are no holes to fill, and the TIM has worse heat conductivity than copper, you'll be negating some of the performance increase of the lap process);

or

2) The TIM has a higher heat conductivity than copper, meaning using it will end up being a waste of money, since, again, there are no holes to fill, and copper can only transfer so much heat; in this case, the heat will not "transfer faster", because of the "lowest common denominator".

Theoretically you are absolutely right. The thing is just that you won't get the surface perfectly clean and flat just by hand lapping. There'll still be microscopic air gaps that you don't even see as scracthes with the bare eye. But under the microscope the surface still looks like a mountain chain. These air gaps will still decrease the performance without any TIM noticeable.


And a TIM will never reach or even surpass the thermal conductivity of a solid metal such as copper. Look at the liquid metal TIMs. They are far superior to any other TIM on the market, but are still hundreds of miles away from reaching copper's thermal conductivity.

Ultimately, this would explain why some people don't see improvements on temperatures after lapping - the TIM holding them back -, actually meaning that, if you "mirror-finish", then you're better off NOT using TIM... Weird, huh?

Does this seem absurd at all? It does make sense to me, but, as I've said, I'm a law major, to the last time I even had a Physics class was like 12 years ago... hehehe

Cheers.

Miguel

Again, theoretically correct, assuming the surface would really be absolutely flat. But in practice it never is. The best you can get by handlapping is no visible scratches and close to a flat surface. Most likely you'll grind down the edges while hand lapping, but that doesn't really mean much for performance, as the CPU DIE is situated directly in the middle.
If you have a shiny and flat surface then using most compounds can indeed decrease performance due to the particle size being too big. For example silicon oil could be used as a TIM here. Many people already did so with good results. But as always this is experimental.


One more thing: Don't forget that there is one more thermal resistance present in the system. It's not only IHS to heatsink, but also CPU DIE to IHS. That's why some people remove the IHS. It's in order to completely eliminate one of the two thermal resitances from the thermal equation.

If the thermal resistance between the IHS and the CPU DIE is quite high (i.e. suboptimal contact) you can lap the IHS to perfection and still you'll get suboptimal thermal performance.



And no, your thoughts don't seem absurd at all. Considering that it's 12 years that you heard the basics, I'm shocked how well you still remember them. I sure wouldn't remember that, if I was in your shoes.