Ivy Bridge Temperatures Could Be Linked To TIM Inside Integrated Heatspreader: Report

[Buckster]

I really hope Haswell has no issues like this...

at a guess this problem won't go away as we shrink further ...

 

[neo-omega]

So when will OcUK be selling "Topless!" Ivybridges then

 

[jeff6475]

at a guess this problem won't go away as we shrink further ...

then lets get intel to make them bigger

 

[the_director]

Interesting results - and many thanks for them - but I can't say I'll be reaching for a sharp knife anytime soon.

 

[mmj_uk]

Has anyone seen or heard of any attempts at solder-mounting the IHS on to the core?

If it worked so well for Sandybridge, why not for Ivy?

It didn't work that well, IHS never do compared to a 'naked' core making direct contact with the heatsink.

Sandy Bridge simply didn't run as hot so the extra solder+IHS layers didn't hinder heat transfer as much.

 

[winter zombie]

thats a good sign

 

[pmc25]

See my earlier post - there was a problem found in Intel's own research with the solder where it suffers stress fractures under repeated heat cycling.

Obviously a fracture = air gap, which is very BAD for heat dissipation. It's thought this is why the change was made.

That was long prior to Sandy Bridge, and there was no problem with Sandy Bridge's solder, in either 1155, 2011 or the bigger Xeon boards. That was long, long ago. They continue to use solder in their IB laptop chips.

 

[hanluc]

I've just done my i7-3770K. This test is with a custom loop, triple RAD, 3 x Gentle Typhoon 1450 RPM on max for both tests.

Here's the before, this is with mx-4 on top of the IHS:

Here's the after, I removed the IHS and applied liquid pro on the die, and also liquid pro on top of the IHS:

 

[sumyungi]

Well done hanluc,

Adding up the averages of the maximum temps, thats a 16C drop in temps

 

[littlepuppy]

Nice work. Can you get 5ghz now?

 

[hanluc]

Nice work. Can you get 5ghz now?

I've not tried, I'll give it a go later

 

[Krono5]

So it is cheap TIM instead of solder then and not the removal of the IHS completely that causes the difference in temps?

 

[hyperst]

Very nice hanluc, that's pretty much conclusive then, Intel messed up.

 

[hanluc]

So it is cheap TIM instead of solder then and not the removal of the IHS completely that causes the difference in temps?

I don't think it's a case of "cheap" TIM, because I tried mx-4 in between the IHS and die and the temps actually went up, it took me a few goes to get it back to what it was before. I rate mx-4 quite highly, I've used it for years, I just think think the combination of paste>metal>paste>metal is acting as an insulator. Coollaboratory liquid pro, is liquid metal and doesn't seem to be affected by this.

 

[Jokester]

Liquid pro isn't anything miracluous, it's only 1-2degC better than normal TIM.

 

[Ollie1132]

Hm now I want to try this but I think I will give it a skip for now. At least until I have the cash for a spare chip incase it goes balls up xD

 

[hanluc]

Hm now I want to try this but I think I will give it a skip for now. At least until I have the cash for a spare chip incase it goes balls up xD

I can tell you when I was pushing the stanley knife blade into the corners of the IHS, a voice in my head kept shouting "WHAT THE HELL ARE YOU DOING!" ... but I'm glad I did.

 

[Jokester]

I can tell you when I was pushing the stanley knife blade into the corners of the IHS, a voice in my head kept shouting "WHAT THE HELL ARE YOU DOING!" ... but I'm glad I did.

Done like a pro

 

[Ollie1132]

I can tell you when I was pushing the stanley knife blade into the corners of the IHS, a voice in my head kept shouting "WHAT THE HELL ARE YOU DOING!" ... but I'm glad I did.

Ahah I'll give it a shot in a year or so when I make my first custom water loop

 

[greywolf]

Tempted to swap my 990x for one of these, as replacing TIM seems to help. So the IHS is fully steel? Might mill a full copper one out and polish it up to see what effect this has.