The problem is, even the most accurate, precise, controlled setup is somewhat individual. Even in an identical test bed your numbers will probably be different. Absolute numbers don't exist and can never exist. The only valid measurements are relative. i.e. relative to this block A, block B is 10.0125C cooler for the same load and coolant temp.
Since B was measured against C in someone elses setup and was 6C cooler, it can be assumed that Block C is 4.0125C better than A. But you have doubts already. Was the water temp and load the same for all measurements? Probably not. Were the thermometers calibrated the same. Almost certainly not. The noise in these measurements already mean that the data is in question and the precision i.e. accuracy to so many decimal places, is useless. The only valid data is that B is 10.0125 C cooler than A in that setup. Assuming that this was a perfectly calibrated setup. You can make assumptions which are reasonable, just not accurate. Player0s assumption that the WW was so many degrees above the RBX was reasonable. Not accurate but about as good as anyone will get without including that block in his tests.
So, the only way to compare is to compare blocks tested in the same setup. Not necessarily at the same time, but you have to trust that conditions don't change. Even if that setup is somewhat uncontrolled (ie water temp varying because it hasn't stabilised for 24 hours, on a CPU instead of a super calibrated resistive heat load) it doesn't matter, as long as it's consistant for each block tested and the variables are measured. Variables can be accounted for mathamatically as long as they are known.
The noise in this data comes from the unmeasureable variables (TIM thickness, mounting accuracy etc) and that is exactly what makes absolute, accurate to 6 decimal places, measurements meaningless.
I personally measure things every day. I am measuring and trying to maintain an accuracy of within 10nm 3 x standard deviation on linewidths of 300nm , using a measuring tool that is repeatable only within 5nm, produced by a tool kept at a temperature within 0.001C of 20C which uses a laser with power fluctuations in the order of 6% (3s). We have to account for these fluctuations because the linewidth will vary by up to 3nm per 1% fluctuation. It is done by monitoring the variables. As I said If they are known, they can be accounted for. These linewidths are nominal, they are only valid against linewidths measured in the same tool, or in a tool calibrated against the same reference.
And that will never happen in the measurement of waterblocks, unless some rather large effort is made to make a standard waterblock type the reference. And even then I think the validity will be lost in the noise of unmeasureable variables.
You won't get absolute numbers.
Cheers
|