Rockhammer, You wrote...
>Agreed to both! Like I said, I don't dispute the fact that socket >thermistors under report CPU temps, but I still contend that as long >as they always under report temps they are valid for basic comparison >purposes. Precision is the key here, not accuracy.
Yes, precision is the key if the "basic comparison" is simply to "rank" the heatsinks or interface materials being tested. But most testers claim that the temperature "differences" reported by socket thermistors are the "same" as what the cpu sees, and that is "not accurate", it's wrong. That's what we hope you got from the DC circuit analog models.
>If we are going to contend that socket thermistors are not a valid >means of testing HSFUs, we must find the mechanism by which one >particular HSFU will cause the thermistor to under report to a >greater degree than another HSFU. Once we?ve identified a potential >mechanism, perhaps we can design an experiment to test the theory.
The mechanism which causes the problem is that the measurement is being made in a secondary heat flow path instead of inside the cpu.
The reported "compression" of cpu temp differences is the nature of the heat flow circuits, it's not a measurement error. It's an error of interpretation by people who do not understand the physics nor the issue.
And the issue is not the "degree of under reporting" of various heatsinks or interfaces. The issue is the "inaccurate" reporting of how much the internal cpu temp changes from one hs or interface to another. Accurate cpu temp change is what's important to overclockers.
For example, many testers using socket thermistoprs report Arctic Silver only improves cpu temp about 1 or 2C, which they say is not enough to justify its cost...and that might be a reasonable opinion if that is all the cpu temp changed. But tests using the "internal" thermal diode of Intel chips show improvement of around 6C.
Which covers your last comment about designing a suitable experiment,
test components and materials on systems running Intel chips with internal thermal diodes. Nevin House of Arctic Silver has done so and we have also co-operated with him in establishing a method for calibrating thermal diodes...though calibration is not essential as they do tell the same lies every day...they are precise even when not accurate, so they do report "accurate differences".
We don't see the data at
www.arcticsilver.com from tests Nevin ran using thermal diodes, but you can write to him from his site. Sure he'd be glad to send you the same results he sent to Mike and us.
(Calibration article is at site).
Hope this clarifies all the issues, and glad to see the knowledgeable interest in them.
John C.