An in-situ performance test of liquid cooling for a server computer system

An IBM power series server ×3350 was tested by using the liquid cooling system developed by Yang et al.. The performance of the heat pipe based direct air cooling system equipped in the original server was also tested for comparison. Both CPU core temperature and cooling system noise level were measured in the present test, he test results of liquid cooling and direct air cooling show that the maximum average core temperature to air temperature difference (at 100% CPU loading condition) is 46.5°C for IBM original direct air cooling but only 24.8°C for liquid cooling at heat sink fan speed of 2200 rpm. At which the noise level is 59.5 and 49.5 dB respectively. While reduced the heat sink fan speed of the liquid cooling system to 1080 rpm, the core to air temperature difference slightly increased to 28.4°C, but the noise level reduced to 36.5 dB. The liquid cooling system required only 1/5 to 1/3.5 of power of the direct air cooling did. These results indicated that the liquid cooling system provides much higher cooling performance but lower noise and lower power consumption than the direct air cooling system for high power server.