Test bench idea....

[maxSaleen]

I have noticed over the past year that a common issue that arises on the forums is the difference between testing with a large die and a small die. The difference being that large dies are used to simulate a CPU with an IHS, and a small die being used to simulate direct WB to CPU contact.

Seeing that (almost) all contemporary CPUs use an IHS, coudn't one find an "optimum" solution by placing a (fabricated) IHS over a small die (thermocouple)? Would this not give the best indication of "real world" usage? Just a thought, and IMHO, obviously, seeing that I'm not a PRO.

[BillA]

not possible
you will be unable to REPLICATE the mfgr's TIM joint
you/we know neither the matls nor the procedure nor can we replicate the (mechanical) mounting with a loose IHS/shim

anything other than a CPU will require correlation
correlation of raw temps to specific CPU case temps has only been done for the heat sources being criticized, and is not public in any case

2 somewhat extreme sizes could be used to look at source size sensitivity ?
or 1 avg ?
lol
and what about rectangular CPUs and wb orientation ??
-> do we have to test every configuration ? (what/who defines the combinations 'needed' ?)

[maxSaleen]

do we have to test every configuration ? (what/who defines the combinations 'needed' ?)

Perhaps that is the purpose of an open forum? The community, through a forum (cough...procooling), can dictate what "standard" testing procedures are. No?

you will be unable to REPLICATE the mfgr's TIM joint
you/we know neither the matls nor the procedure nor can we replicate the (mechanical) mounting with a loose IHS/shim

Perhaps Intel and/or AMD would be nice enough to tell us how they mount the IHS. Can't imagine that is very sensitive material.

anything other than a CPU will require correlation

I was going to mention that the suggested device be caliberated to different CPU speeds. Let's just say:

Athlon 64 venice @ 1.6ghz= 50w
@ 1.8ghz= 65w
@ 2.0ghz= 80w (all made up off the top of my head)
The device would have to be caliberated (fiddled with) until its temps matched that of a CPU putting out the same amount of heat.

This is kind of like a dog chasing its tail..... ahh screw it. I knew this idea was out of my leauge of thinking....