I could be completely wrong with any of these, as I'm basically in a state of learning disability, but....
What I was thinking is to have a system, with WAY too many heatercores, say 2-3+ in parallel. I think this would take *most* of the heatercore's performance in the system out of the equation. This would reduce overall flow restriction through the HC part of system, and since the water would be flowign thru each HC a lot slower, it should have no problem cooling the water to room temperature(which would have to be kept constant).
I've seen " Y " bypass valves that switch flow from one side to the other. You could have one line going to the waterblock and the other side going to a bypass line that just goes through a 2nd block of the exact same brand/type with the same exact length of hose. This bypass line block wouldn't be hooked up to a heat load, it would only be there to keep the same flow resistance on the 2nd line. As you switch from one line to the other, the pump wouldn't see any difference in flow resistance. This would prevent fluctuations in the temperature generated from the pump itself. You would also need a "dummy" flow meter on this line as well, just to keep any flow restriction caused by a flowmeter equal on both lines.
Now, with any worries of cooling system inadequacies and pump temp fluctuations removed, you can focus on the block itself. You would now adjust the Y bypass valve to raise/lower the flowrate through the testing block with a constant heat load on it. Start off with low flow rate, and measure temps as you
slowly sweep through to maximum flow rate, letting the system fully stabilize at each recorded point.
As for mounting the blocks, you could do either or both of the following:
1. Use the hardware included with the block, and follow manufacturers directions on mounting.
2. Mount all blocks with the exact same pressure on the cpu die. This would require a calibrated precision torque driver following proper torquing procedures to torque the 4 bolts to a set amount.
If I'm thinking correctly, all this would help isolate the performance of the system to the block design itself. You should be able to get somewhat accurate results of the performance of the block itself, then plot it for comparison to other blocks.
************************************************** *
Now if I'm missing something (which I probably am) I hope these ideas can at least be tweaked/improved upon for better results
I'm not even going to try to get as detailed in procedures as BillA would as I'm just not as edumacated
