Pro/Forums - Does this make any sense to you? Unregistered? pHaestus? Les?

Pro/Forums (http://forums.procooling.com/vbb/index.php)

- Testing and Benchmarking (http://forums.procooling.com/vbb/forumdisplay.php?f=61)

- - Does this make any sense to you? Unregistered? pHaestus? Les? (http://forums.procooling.com/vbb/showthread.php?t=6388)

Does this make any sense to you? Unregistered? pHaestus? Les?

Quote:

Originally posted by Someone
as for acurate, I mean by +/-1c wich is really good. I use an 8k3a with the on die probe, air probe, and water probe, I graphed and made a linear regression modle to see that my ondie probe is only .1c off.

Quote:

Originally posted by Someone
yup its a pally heres how I did mine

1400mhz 1.75v 63watts

35, 24.0----------------------11

1750mhz 1.8v 83watts

38-39c, 24.2 -----------------14.3

1818mhz 1.85v 91watts

40 24.2----------------------15.7

1848mhz 2v 109watts

44, 25.0----------------------19

the numbers on the side were the delta between water and cpu, then I took the wattages, at different speeds. I plotted them in xstat and ystat on my ti86, then I made a linear regression model and checked out the graph and the table. at 0, it should line up with 0, make sure you have it correctly, at first i thought my temps were off by 3c, but it turns out I had the numbers flipped around.

put the delta in the xstat and the watts in the ystat, then go linR xstat, ystat and hit enter

It makes sense. Can't say whether or not it's actually correct.

He's assuming the heat transfer coefficient between the die and the water in the block is constant (probably true) and making a graph of q vs. deltaT. This should be a linear correlation, or nearly so, and should have an intersection point very near zero.

The problem is that he assumes he can get a temperature accuracy of the water in the block much better than 0.1C. Not likely.

This sort of technique is pretty common in chemistry and heat transfer problems, but I more often use MS Excel since I can type much faster that way.

I think it would be a good idea if done by someone who includes propogation of error due to instruments.

Alchemy

Maybe I looked at this too quickly but aren't you just looking at linearity not accuracy. Over that small of a range I would expect that kind of linearity.

But, I'm probably off in left field somewhere.

http://www.jr001b4751.pwp.blueyonder.co.uk/Geek.jpg
The data suggests a linear relationship betwwen the measured quantities. However it is dangerous to extrapolate.
Gives no information on the relation between "Degrees Geek" and "Degrees Centigrade", or the accuracy of the Wattage calculations.
Is useless for calibration purposes.