windows - Heat and its relationship to Frequency in overclocking

I have a i7-4770k - In order to Overclock it from 3.5-3.9(boost) to 4.5 I have to raise the Voltage to 1.387Vcore. This is not an issue with heat as my water-cooling keeps the CPU at 70c under P95 load and in Benchmarks can reach 90c. My question is, if the thermal paste joining the 'Lid" and the actual chip is sub par will that cause the CPU to need to be clocked higher to get a stable clock. Or would that only effect the heat distribution to the 'Lid" from the chip?

Yes because the haswell will fail if it isnt cooled nicely when your really pushing the clock, it is possible that a poor thermal interface from the chip to that first copper heat spreader lid thing could decrease your ability to reach higher frequencies, including all the extra power you end up pushing into it to try and get the transisters to switch reliably at those frequencies.

That is sorta what you asked.

From my experience and from reading about a lot of other users making it to 4.6 with haswell is luck and really good stuff, and high voltages and people making it are really good at what they do, and have really primo cooling, and will even de-lid. Even then they might not use that 24-7.

I think here (not at overclocking land) I can safely say that the huge numbers you read are fleeting glimpses of short term heroics and luckey chip finders.

That lid makes for a really nice heat spreader at first, and although intels thermal contact is not the greatest in the world, trying to fix it will not change things That much. My picture here was typical of the thermal interface for other haswell de-lids that were shown on the web.

The lid is thick and all copper.

The chip under there is a thin (.5mm) easily breakable chip, the lid making it almost impervious to physical damages.

I did de-lid mine to try and fix my cooling problems, but changing my cooler did more good than de-lidding. And de-lidding came with a lot of other possible things that could go wrong. (other than the loss of warrenty)

With my cooler the clamp had to go

The other componentes had to be protected from goop
I had to add pads to keep from chipping the edges.

I had to completly flatten the heatsync, which would have busted the cpu chip easily because it was convex (bulged) for a lid.

These socket pins bend if you look at them wrong :-) and (of course) are critical.

Remember the haswell is a percentage faster than the previous generation at doing stuff (general) so even if you do not achieve vast pretty numbers and are lower looking and actually 100% stable, your still going faster :-) How many people have wasted buku time and money trying to reach the sky, when they were already in the sky (because the chip is faster). Get your fully reliable lots extra , and go on with life , the last 2% ain't worth it.

De-lidding has changed cooling for a lot of people, but I have not seen a lot of it completly changed the overclocking ability itself. It changed my cooling, but my overclock max did not change.