1) To be useful, good numbers for flow rate and wb inlet and outlet are needed at the minimum. Drilling out baseplates and getting a bp temp would be suggested as well. Dunno how useful CPU temps are yet; playing around still with better gear.
Here's what I use currently:
For water temps (wb in and out) and air temps (rad in and out) I am using a Digitec 5810 thermometer and YSI 700 series linear thermistors. Not perfect, but two of the water probes read exactly the same (to 0.01C) over the range of water used.
For CPU temps I use a Maxim 6655EV System from dallas semiconductor. I bought it because the homemade readers are limited to the resolution of MBM (1C) while the maxim software is 0.125C (over || port). It took some fiddling to get it working acceptably.
For wb testing, I am primarily relying upon the delta T across the waterblock at this time as I still don't know how relevant the diode temperature of the CPU is. I am pretty confident a superior block will put more heat into the water though.
Flow rates are basically "the master variable" for wb testing, and I use a GPI paddlewheel type digital flowmeter. It has a substantial pressure drop, and I compensate with a beefy pump (Little Giant 3E-12NYS)
All this was ebay/surplus/leftovers from BillA. Not sure about the YSI 70x thermistors, but there are some cheap digitecs on ebay now.
re: 1-2C error. We have gone over this before; start propagating the error forward in the calculations and you will see why better numbers are needed (Uncertainty proportional to sq root of the sum of squares).
2) Maybe we could get Bill to put some of his tests with Swiftech bps of differing thickness and material? I think he has changed his testing methods since the testing was done, but he might be convinced
3) Also cheaper just to buy a faster CPU (at least with AMDs). Stock CPUs run 60C no prob; why not just stick with whatever free MHz you can get with the retail heatsink and thermal pad?
|