SB to IB - Intel cheaping out on thermal solution

[MrPickins]

Intel doesn't want to neglect the enthusiast market which you guys seem to be part of. Communication is key.

The enthusiast market is nowhere near as important as the mainstream market, who have no idea that there is an "issue".

Why would they make a statement that brings it to the attention of the masses?

 

[dave_the_nerd]

Intel doesn't want to neglect the enthusiast market which you guys seem to be part of. Communication is key.

Maybe so, but as long as they have the fastest chips and/or the best product, the enthusiasts will keep snapping them up - even if our rigs' load temps are +5˚C because they cheaped out.

 

[Magic Carpet]

The enthusiast market is nowhere near as important as the mainstream market, who have no idea that there is an "issue".

An issue? I'd rather call that a design feature. Nobody exactly knows why Intel did that. They surely had known the consequences it might lead to.

Why would they make a statement that brings it to the attention of the masses?

Again, it would make good PR and show that they care. But instead, they deemed it not worthy their attention, which shows their neglect towards their customers (albeit a very small %). The Internet is plagued with speculation, benchmarks and conclusions by "worried" enthusiasts. It's natural for good business to come forward and explain clients' "frustration".

 

[Phynaz]

Intel doesn't want to neglect the enthusiast market which you guys seem to be part of. Communication is key.

Except it isn't an issue. So why would they comment on something that doesn't exist?

 

[blastingcap]

So I've heard a lot of talk about how Intel cheaped out on the thermal solution when moving to Ivy Bridge from Sandy Bridge.

What exactly is the difference? Can someone please clarify?

I'm assuming it's the following:

SB: Soldered interface between the heat spreader and the CPU die.
IB: Thermal interface material/pad between the heat spreader and the CPU die.

Do I have the right idea here?

Idontcare has done a ton of tests with all sorts of thermal pastes etc. and IIRC he believes that Intel didn't cheap out as much as you might think.

 

[bononos]

An issue? I'd rather call that a design feature. Nobody exactly knows why Intel did that. They surely had known the consequences it might lead to.

Again, it would make good PR and show that they care. But instead, they deemed it not worthy their attention, which shows their neglect towards their customers (albeit a very small %). The Internet is plagued with speculation, benchmarks and conclusions by "worried" enthusiasts. It's natural for good business to come forward and explain clients' "frustration".

I think someone (IDC) mentioned the possibility of thermal paste accommodating extra flex which plagued Nvidia in the bumpgate incident.

Its way more frustrating on the other team when their cpus could suck down 200W despite being spec'd as a 125W tdp.

 

[jpiniero]

Maybe so, but as long as they have the fastest chips and/or the best product, the enthusiasts will keep snapping them up - even if our rigs' load temps are +5˚C because they cheaped out.

I was under the impression that Intel went to TIM more because it better handles the heat density that shrinking to 22nm brings. How long until Core is at "Rocket Nozzle"?

 

[Idontcare]

Idontcare has done a ton of tests with all sorts of thermal pastes etc. and IIRC he believes that Intel didn't cheap out as much as you might think.

Yeah the TIM itself that Intel uses is at least every bit as good as NT-H1 (a premium TIM in its own right).

The problem with using the TIM instead of the solder is the overall height of the gap that exists between the CPU and the underside of the IHS.

If Intel filled that gap with solder instead of TIM then the gap isn't a problem. But filling that large of a gap with any non-solder TIM is going to cause the thermal bottlenecking that we see with IB, its not a matter of using expensive or cheap TIM at that point.

Delidding pays off because we reduce the gap in the process of replacing the stock TIM.

I think someone (IDC) mentioned the possibility of thermal paste accommodating extra flex which plagued Nvidia in the bumpgate incident.

Its way more frustrating on the other team when their cpus could suck down 200W despite being spec'd as a 125W tdp.

As an engineer that is far and away my top concern with delidding. There are mechanical forces in play because of thermal expansion and the mismatch in materials (silicon's coefficient of thermal expansion versus that of copper).

It is not unreasonable to suspect that Intel's engineers steered clear of the rigid solder and instead opted to adopt the more flexible thermal pad intermediate entirely out of concern over the cyclical thermomechanical stresses that are generated in heating and cooling of the CPU die.

But we cannot be certain either. The mechanical forces are surely present, but they may not be large enough to be a practical concern. And the decision to avoid solder may have purely been one of cost-reduction, in which case we enthusiasts are not compromising anything in delidding as it were.

 

[Borealis7]

Intel SHOULD keep us Enthusiasts happy, because eventually we recommend other non-tech-savvy people what hardware to buy. it's time we get what WE want.

power to the 1%!

 

[Phynaz]

Intel SHOULD keep us Enthusiasts happy, because eventually we recommend other non-tech-savvy people what hardware to buy. it's time we get what WE want.

power to the 1%!

Yes, everyone's CPU lifetime should be reduced to keep you happy 🙄

 

[crashtech]

Makes me wonder if the correct thickness of copper shim coated on both side with a good TIM would help for those who want to re-install the lid.

 

[Ken g6]

Makes me wonder if the correct thickness of copper shim coated on both side with a good TIM would help for those who want to re-install the lid.

Three layers of TIM? D: Maybe if you soldered the copper shim to the underside of the lid...

 

[crashtech]

Stock, there are two, and apparently the one under the lid is of excessive thickness. Anything to reduce that thickness with something more conductive is a win, it's not as simple as counting layers.

 

[Idontcare]

Makes me wonder if the correct thickness of copper shim coated on both side with a good TIM would help for those who want to re-install the lid.

The shim is not necessary. Simply removing the existing TIM and replacing with any TIM which is more pliable than the stock TIM when the IHS is placed back on top of the CPU is sufficient.

 

[crashtech]

The shim is not necessary. Simply removing the existing TIM and replacing with any TIM which is more pliable than the stock TIM when the IHS is placed back on top of the CPU is sufficient.

Has anyone attempted to measure the clearance between the die and spreader? Of course it would vary from unit to unit, but I was curious if some nominal value had been established.

 

[2is]

not only poor heat transfer is the culprit, but IB has higher transistor density than SB (due to the move to 22nm) but roughly the same die area as SB and that increases the heat output even more.

The heat output isn't more, it's less. Operating temp != heat

 

[Phynaz]

Has anyone attempted to measure the clearance between the die and spreader? Of course it would vary from unit to unit, but I was curious if some nominal value had been established.

Go find IDC's thread about delidding his 3770K. Everything you want to know is there Grasshopper.

 

[Idontcare]

Has anyone attempted to measure the clearance between the die and spreader? Of course it would vary from unit to unit, but I was curious if some nominal value had been established.

Checkout this thread, pay close attention to all the linked posts captured in the OP.