[Cathar]

Originally Posted by unregistered

Cathar
the term is 'case to air', where 'case' is the IHS
nothing at all to do with the silicon temp
I was not attempting to make any characterization of the silicon temp
this activity is to characterize thermal resistance where the transfer is between the IHS and the heatsink, and the IHS heating is done by a device designed to do so

Such does not characterise actual die-temperature cooling efficacy if there are unknown and unquantified variations between the die heater and the IHS with each IHS-cooling device mating.

I believe the device may be considered to be more representative than a correction factor applied to a different device - speculation on speculation here

Representative of what though? What is being measured? I mean we know what is being measured, the IHS surface - air delta. I still fail to see how that helps in allowing for an accurate determination of CPU die cooling efficacy when the CPU die to IHS contact is varying on a per cooling device basis.

An Intel TTV is fine for heatsink validation, for providing a cursory and repeatable "pass or fail" assessment of heatsink solutions. It is not representative of actual real-system cooling efficacy, and not even Intel treats it as holding that level of determination. It is merely a validation tool, not a final performance measuring device.

your concern would be better placed with the 'size correction factor', how will you determine such ?

Meaning the size of the heater-die? I find such a statement a little ironic when the size of the heater-die in an Intel TTV is neither known nor documented.

what is the baseline ?

Something to be determined perhaps? Something to be independenly assessed through trial and error and actually measuring the temperature of the heater-die/CPU-die simulator, rather than the outside case/IHS temp.

Ben's WTBA ?

Please. I was thinking more along the lines of Incoherent's excellent work for much of the above.

every time I have the opportunity to follow 'industry practice' I do so, avoids time in discussion and cross-testing (and my ongoing self education)

...otherwise known as following the path of least resistance through accepting the lowest common denominator, regardless of suitability for the intended purpose.

There are bare-die CPU's from both Intel and AMD aplenty. There are bare-die GPU's almost solely. The variances in CPU die->IHS contact are well known and documented and observed far and wide. How are such catered for with an IHS-only unquantified-variable-laden testing methodology?

It would seem to me that insistance upon adherence to a particular testing solution which is both clearly flawed and even acknowledged by the testing solution's manufacturer as being inadequate for purposes of final performance assessment is purely a matter of convenience and little else.

I would say that everyone absolutely loved your work with your old thermal testing setups, not because they were "standard", but because they challenged people to do things correctly, to know and quantify the variables, and sought to explore and understand the very nature of what it is that we were all concerned about measuring, being the die temperatures and effects of mounting variations and TIM layers. I see using a "standard" TTV where at least 3 variables are not even known, and worse, completely ignored, as turning one's back on the good work and progress laid down before.

I, for one, don't understand it...