- Lian Li PC75 USB
- Asus P4G8X
- 2x256MB Corsair TWINX 3200
- Intel Pentium 4 3.06
- Sapphire Radeon 9500 128
We ran a set of SiSoft Sandra Burn-in wizards, ten loops after having used the setup for three days (to make sure the thermal paste had "settled"). The looping was performed three times and the highest temperature was registered as result. The idle temperature was measured in the morning after, before any work was done. Ambient temperature fluctuation was +/- 0.2 degrees. The same procedure was performed after overclocking to 3450MHz.
P4 3.06GHz at default speed
As we expected, the Hydrocool ended up below our other watercooling systems because the AC and Magnum waterblocks were tested using a much bigger and more efficient radiator, pump and so on. Despite this, the Hydrocool performed very well - the DeltaT between idle and full load is only 3.5 degrees. But what happens when we increase the load on the cooler?
Overclocked, P4 3.06 running at 3450 MHz
Overclocked to 3450MHz, the idle temperature increased by 7 degrees and the load temperature by 5.5 degrees, ie. a DeltaT of 2 degrees between idle and load.
The performance in turbo mode didn't improve the performance much -- 0.5 to 1 degree cooler, so there's really no reason why one would want to run the fan at full throttle as the noise level tends to be a bit annoying.
After the testing, I disassembled the kit. What we found was quite disturbing -- the processor made very poor contact with the waterblock, and the reason? The waterblock is concave, making the contact-area between the two surfaces shrink to only a fraction of what it should be.
Poor contact with the processor, the waterblock base is concave
No wonder the Hydrocool 200 doesn't cool as well as expected with that poor contact between the processor and waterblock. As usual, we used a very thin layer of thermal paste, and we always clean off the processor before mounting any new cooling equipment. The image above speaks for itself -- only the edges of the processor had made contact with the waterblock, and we have verified that it is the waterblock, not the processor that is concave.