There is hope that for Ivy Bridge E Intel will use solder, since that will be a larger die.
But the die area occupied by the iGPU (in IB and HW) is about the same as the area that would be occupied by
2 extra cores and the extra cache, in IB-E. So it's possible that Intel will use TIM for IB-E, which would suck.
If the 6 cores in IB-E are 6 out of 8 and cache is 15 out of 20 then Intel will definitely use solder IMO.
If a pure 6 cores die will be more square-ish Intel could use solder and leave some spare space in the corners for the cracks.
The hottest research topic in CPUs should be finding a better solder, one that doesn't stress / crack the die.

It's not just CPU's affected by this problem (BTW its the solder that forms cracks not the die)

Since the European union banned the use of lead based solder in 2006 a lot of research has been going on.
The problem of solder forming cracks/voids is a big problem for the whole industry, along with how to remove high heat loads from ever shrinking devices.
Everything is being looked at to solve the problem including Carbon nanotubes, Gallium based solder, and making the join in a vacuum just to name just a few.

Personally i think Intel using standard goop between the die and the IHS is a stop gap measure until something better is developed.

It would seem your speculations about "hope that for Ivy Bridge E Intel will use solder, since that will be a larger die." maybe true Tzupy
http://www.hardcoreware.net/solder-ivy-bridge-e-4960x-delidded/

http://www.youtube.com/watch?v=2eFzNpTOaOw

^If you're feeling brave and don't have the patience for a razor...

I delidded my Haswell using this method -- don't worry, unlike the guy in the video, I kept my shirt on

It helped drop temps by about 15-20C at Prime95 load (from 95C to 70C using a Prolimatech Megahalems and a Scythe EX-FDB fan). That being said, load temps are a lot more variable on Haswell than on my last processor (Nehalem-based i7-920). On the 920, load temp was load temp. Firing up Prime95 and starting any FFT size would load my proc to about the same temp. With Haswell and AVX, a big FFT that doesn't fit in the caches (and presumably spends a lot of time accessing system memory) will only put the core temps around 75C. A small FFT size (e.g. 8K) shoots my temps all the way to 98C.

I hope you mean 98C before delidding.
Would you mind telling us which TIM you used, maybe CLU?

Saw that delidded IB-E and wanted to post a link here, but Spoon Boy was faster.

CPU | Die Size


If that 2006 article by Intel is right, then does that mean they used thermal paste in the dual-core Sandy Bridge CPUs (i3 2xxx, Pentium G8xx/6xx etc.)? Anyone had the chance to de-lid one of them? I gather it would've gone under the radar since they can't be overclocked.

Unfortunately, no. That 98C is after delidding. TIM is Tuniq TX-4. Again, that temp is only when using AVX extensions with small FFTs in Prime95. Real full-throttle loads only go to about 75C.

Another thing is my chip is an absolute dog when it comes to overclocking (meaning really bad). I need 1.35v just at 4.2GHz. From what I hear, it's not unreasonable to get 4.6GHz @ 1.2v. So that's where a lot of my bad temps are coming from methinks.