determining overall C/W with idle/load temps?
 

ok, so motherboard measurements are terrible for ABSOLUTE temps

but, they seem to be decent for relative temps...that is, if you increase the environment 3C. they increase about 3C.

Would it be unreasonable to measure the difference between idle/load temps, and somehow calculate the difference in watts of heat output of the CPU?

I think your first assumption is incorrect. All the boards I have used do not change 3C for 3C. For example my Abit KD7 will be fairly linear to 40C. Then for every C the thermocouple pics up the on-board pics up .5C and it gets worst the hotter the CPU gets. I seen the same on my ASUS A7V333, A7V8X-X, Epox 8K7A (X5), ECS K7s5a and Abit KT7A. I have yet to have a board read properly once it gets over 40C.

If you are planning on doing this then try and calibrate the on-board with a TC or something "good" or you may spend a lot of time collecting useless data.

Note this is my own half assed testing so who knows. Better safe than sorry though.

There's plenty of evidence in addition to Jaydee's that onboard sensors are distinctly non-linear.

I sit corrected, then. Just thought I'd throw it out

funny ABIT RD is just looking to electrical resistance vs. temp past week, I guess someone in their BIOS team finally realize that it is NOT linear but a curverture relationship