I think its because the IHS is not perfectly flat (usually concave). I know a popular thing to do with waterblocks is to bow the surface with thicker washers.
Well anyone lapping their heatsink will probably lap the IHS of their cpu too. The thing is, while the line down the middle is slightly concave, the TRUE is just too convex to make proper contact. This can be verified by removing the HS after mounting.
Don't forget, metal will still flex under intense mounting pressure. A little more convex is better than concave surface without contact on the most important area, the center. That's why manufacturer has started making convex surfaces. Swiftech, Thermalright were very famous for excellent finishing. Why would they do that if it wasn't a good idea?
Why convex? The answer is very simple: to increase the mounting pressure at the center of IHS. (which is the hottest spot on the processor) Let's compare convex base vs flat base (let's just assume the IHS is flat or just slightly uneven). By making the base convex, you are basically reducing the area of contact between the base and the IHS. Does this mean you will get worse temp? Nope, not really. Because: Pressure = Force / Area Assume the same amount of force is coming from the mounting mechanism for both convex base and flat base waterblock/heatsink. By reducing the area of contact, you are essentially increasing the amount of pressure applied on the IHS, which is concentrated at the center of the IHS. This effectively lowers the thermal resistance between the base/TIM/IHS. This is why a bowed Apogee GT outperforms a flat Apogee GT waterblock easily and by a significant margin, one of the many examples.
If you actually applied THAT much mounting pressure to the point that it causes the copper base to flex, it won't return to it's original shape anymore because copper has quite low yield strength (not sure if this is the correct term). If that happens, you won't be able to use the waterblock anymore. I don't think it's going to happen on heatsink base because they are just too thick, no matter how hard you try. The bases (particularly waterblocks) are usually thick enough and shouldn't flex if you just hand tighten the nuts + bolts, even if you try really hard.
So yeah. I are teh slow. 3.8GHz on stock 1.2v (still finding stock voltage baseline). CPU-Z Validator 2.1 8 hour prime stable (I also used IntelBurnTest as a quick acid test). Still pushing (and loving it).
4.2GHz (1.32v) and counting! 20 Passes IntelBurnTest (I will save a big prime95 for last). CPU-Z Validator 2.1
4.3GHz (1.35v). IntelBurnTest cooks it to 71C but Prime Small FFTs only manages 63C. CPU-Z Validator 2.1 Might need to stop due to temperatures though...
Stock volts and keep temps low? Well here is some 'traditional' stability. Nothing too surprising as it did pass the IntelBurnTest. HMM! Maybe I should push it just a tad to see what it has left to give...
It bakes at that high speed. I am still messing around with it to see if I can do anything about it (drop the vcore and see) but as is, the raw OC itself really heats the CPU up. Load on Prime is 65C. Load on IntelBurnTest is low to mid 70s. 10s range for SP1mb...
I can't understand one thing. My processor is E 4600 Core 2 Duo @ 2.4Ghz. So the latest version of cpuz shows core speed 2400.0mhz. . But your processor is Core 2 Duo E 8400 @3.0Ghz. So cpuz should show the core speed around 3000Mhz. Why the core speed is shown to be around 4000 Mhz?