Excellent discussion topic and glad my past hunches
are finally confirmed and documented!
One thing I disagree with is this section
Concavity of the IHS is a less critical factor with smaller CPUs, like the socket 1150, 1155 or 1156 CPUs. The dimensions of the IHS surface being smaller, the same degree of curvature leads to a smaller gap between IHS and heatsink base.
I would argue that it would be even more pronounced. When we first started de-lidding Ivy Bridge CPUs it immediately jumped out at me that the glue might be lifting the IHS too far from the core (link
), and it was confirmed
in Idontcare's extensive delidding thread
If the concavity of the heatspreader can have such a huge effect on a soldered IHS like your 965, on one that is merely secured by TIM & silicone glue, like the more recent 1155 and 1150 CPUs, I feel it would be even more susceptible to deformation.
I think there are other issues with 1155 that complicate the matter -- the stock TIM for one. If the IHS is not making firm contact with the die, then a convex, high-pressure heatsink would bend it, and improve cooling, I agree. Hard for me to contribute much about cooling 1155 Ivy Bridge CPUs, as my experience is more limited. But I will say this about the Intel Core i5-2400 Sandy Bridge CPU on our small HS test platform: Its IHS is slightly convex.
If heatsinks are deforming the IHS into (or further into) a concave shape, what's being equally deformed inside the package? The indium solder does deform over temp (that's part of the purpose - to be the middleman between the IHS's coefficient of thermal expansion and the die's)...but I can't imagine it would be visibly noticable/take the brunt of it. Seems to me the weakest link is the package substrate. Are these things bowing outward? If you look at a CPU with a formerly flat IHS and is now concave, is the backside (pin side) of the CPU equally convex?
This is a pertinent point, and I have to admit the last time I examined our i7-965E, I could not help notice that the corners of the PCB (substrate) are bowing upwards visibly. The subtrate, looked at from the IHS side, is concave! BUT
-- you have to remember that the LGA1366 mechanism pulls the CPU down into the socket by the "lips" on two opposite edges of the IHS and locks it down pretty tightly. The IHS, which is stiffest around its perimeter, is pressed hard against the subtrate when the CPU is mounted, so I'm fairly certain the PCB actually gets straightened. Even if it doesn't, I doubt that PCB bowing has any impact on IHS/heatsink-base contact -- or on die/IHS contact. What impact it has on the longevity of our CPU I'd hate to think. (knocks on wood.) Ditto the 1155 CPU/socket btw. (see pic below of our i7-965E in the 1366 socket).