Testbed and Methods

We assembled this testbed to check the efficiency of the water-cooling systems under question:

• Intel i955X mainboard
• Intel Pentium Extreme Edition 3.2GHz CPU (on Smithfield core)
• 2 x 512MB Kingston KHX5400D2/512 DDR2 SDRAM
• ATI Radeon X850 XT Platinum Edition 256MB graphics card
• Western Digital WD1600JB hard disk drive
• Zippy HP2-6460P power supply

There were two test modes: Idle and Burn (with two copies of the CPU Burn utility running).

Intel’s Desktop Utilities version 2.1, supplied with the mainboard, was used to read the CPU temperature. The ambient temperature remained constant at 18°C throughout the tests. We used the same thermal paste for all the water-cooling systems and filled them up with ordinary drinking water.

I didn’t check the noise level because the WaterChill KT03-L20 Entry comes without any fan, so there can’t be any talk about noise, and as for the pumps, they can be absolutely quiet in both systems if you install them right. As you’ll see below, the resulting level of noise form the system will actually depend on what the user demands.

The system must be assembled and filled up before the tests. In this case, I only have to connect the components with pieces of pipe. The only more or less difficult procedure is water-block installation. The process can be divided into two steps. First, you should install special poles:

The poles come with nuts that prevent damage to the mainboard. And then you put the water-block down on the poles and fasten the screws:

The screws are all fixed at the same height and the pressure on the water-block comes from the springs the screws press down upon. This is a simple and reliable design. It was more difficult with the water-block from the WaterChill KT03-L20 Entry because it didn’t have fasteners for the test platform.

I had to improvise and spend some time to press the water-block down normally, but without any big troubles. I also used one of the fans included with the Antarctica KT03A-12VX 1/2 inch to blow at the radiator.