Lapped my i7-3770K and there was zero improvement in operating temperatures!?

[BonzaiDuck]

I'm not saying you can't lower temps. I'm saying this isn't some great, previously unknown discovery, and Intel isn't going to do anything about it. Anyone thinking they will is deluding themselves IMO.

[It's a dead horse . . . ] I was answering your post and the cat's. . . . :biggrin:

 

[OCNewbie]

fun thread!

 

[Denithor]

I did that with my video card (see thread here) and have every intention of making a decent attempt of it with my 3770k. Just not right away, I want to get some other tests done first because there is always a chance of destroying my CPU with direct-contact testing and if that happens I want to at least have all the other data beforehand.

I might try some indigo extreme, but that is pretty much an irreversible situation given that the bare CPU die itself is involved.

When I used indigo extreme before on my 2600k I had to re-lap both the IHS and my NH-D14 in order to fully remove it all. Removing it from the CPU chip would most likely result in catastrophic failure, so if I put it on there its going to be the very last test I do because it will be permanent.

So basically you'll have to try bare core -> HSF first, assuming you don't break anything then you'd try the IE under the IHS (and hopefully also between IHS/HSF also to give the best straight through connection possible).

This is awesome work, looking forward to more results!

 

[JoeyP]

Is lapping that far (to the point of a mirror finish) necessary? It would seem to me that having some texture would, at a microscopic level, act a little like heat spreaders, and expose more heat to the compound and the heat sink.

 

[BonzaiDuck]

Is lapping that far (to the point of a mirror finish) necessary? It would seem to me that having some texture would, at a microscopic level, act a little like heat spreaders, and expose more heat to the compound and the heat sink.

I always thought that a mirror finish was a bit silly for the extra work. But I can't say those who pursue it with relentless determination are being silly.

A piece of 400-grit or 600-grit wet-or-dri sandpaper will begin to have the polishing properties of a finer grit paper as it is progressively used.

Depending on the TIM, small scratches or a "brushed" finish will trap TIM particulates and hold them in place.

I always stopped at 600-grit, and a finish that left my reflection vague enough to be "almost anonymous." The lapping worked anyway.

 

[Idontcare]

Is lapping that far (to the point of a mirror finish) necessary? It would seem to me that having some texture would, at a microscopic level, act a little like heat spreaders, and expose more heat to the compound and the heat sink.

The benefits from lapping are from the leveling effect, not the polishing or removal of micro-ridges and scratches.

Getting rid of the big bowl that pools TIM in the middle is the benefit, going beyond that is neither hurting nor helping at a practical level. You could probably stop with 220 or 400 grit and realize 99% of the benefit of lapping to 3000 grit.

I go to 3000grit because it is easy to do (takes all of 15 minutes once you have the supplies assembled) and I personally like the aesthetics of a mirror-polished surface.

At the academic level, the maximum thermal conductivity to be had will come only when phonon-phonon coupling is optimally maximized and phonon-scattering is minimized across each interface in the stack.

Ridges constrict the accessible manifold of phonon wavelengths owing to the geometric boundary conditions each ridge creates...the ideal boundary condition for maximum phonon accessibility only comes when the entire surface is atomically flat (no ridges) and with zero point defects (thermodynamically this is only possible at absolute zero for ordered crystals).

 

[BonzaiDuck]

You are a fountain of knowledge . . . .

 

[RussianSensation]

IDC's knowledge is second to none on our forum.

Most interesting to me is if that delidded CPU can actually overclock higher. I wouldn't really care if my CPU ran at 75*C or 60*C but if it could overclock much higher while running at 75*C (because the "new" 60*C CPU has 15*C extra headroom to deal with extra voltage and heat increases), then the delidding would be worth the pay off.

Maybe Intel will fix the TIM/IHS issue that has plagued IVB chips and magically Haswell will be a much better overclocker on the same 22nm node.

Great thread so far and pics IDC!

 

[Ferzerp]

IDC's knowledge is second to none on our forum.

Most interesting to me is if that delidded CPU can actually overclock higher. I wouldn't really care if my CPU ran at 75*C or 60*C but if it could overclock much higher while running at 60*C, then the delidding would be worth the pay off.

Maybe Intel will fix the TIM/IHS issue that has plagued IVB chips and magically Haswell will be a much better overclocker on the same 22nm node.

Great thread so far and pics IDC!

I dropped mine somewhere in the 10-15C range (didn't take good records), so in theory that would allow for a higher OC, but my particular chip is stable around 1.20-1.23V for 4.4Ghz, and then requires massive increases anywhere from there. (I can do 4.7Ghz, but it requires closer to 1.4V), so I just left it at the same speed.

 

[Ben90]

Most interesting to me is if that delidded CPU can actually overclock higher. I wouldn't really care if my CPU ran at 75*C or 60*C but if it could overclock much higher while running at 75*C (because the "new" 60*C CPU has 15*C extra headroom to deal with extra voltage and heat increases), then the delidding would be worth the pay off.

Maybe Intel will fix the TIM/IHS issue that has plagued IVB chips and magically Haswell will be a much better overclocker on the same 22nm node.

Interestingly enough, by making delidding less risky, this is better for the true enthusiasts. How many people delidded Sandy Bridge or Nehalem? It has never been a secret that the IHS just impedes thermal transfer. I see this as a win win.

Checkmate conspiracists.

 

[Idontcare]

So basically you'll have to try bare core -> HSF first, assuming you don't break anything then you'd try the IE under the IHS (and hopefully also between IHS/HSF also to give the best straight through connection possible).

This is awesome work, looking forward to more results!

IDC's knowledge is second to none on our forum.

Most interesting to me is if that delidded CPU can actually overclock higher. I wouldn't really care if my CPU ran at 75*C or 60*C but if it could overclock much higher while running at 75*C (because the "new" 60*C CPU has 15*C extra headroom to deal with extra voltage and heat increases), then the delidding would be worth the pay off.

Maybe Intel will fix the TIM/IHS issue that has plagued IVB chips and magically Haswell will be a much better overclocker on the same 22nm node.

Great thread so far and pics IDC!

Thanks for the kind words and props everyone :$

I went ahead and created a new thread dedicated to the delidding tests and results since this thread was really only ever intended to be about my lackluster lapping results. I've got more to add to the new delidding thread, but its probably a bit of a snoozer owing to its length already so I'll let folks digest it for now before uploading more to it.