Testbed and Methods

The COOL by Corsair kit leaves a highly pleasing first impression. The system features an impressive exterior, uses components from leading manufacturers and experience of Swiftech, one of the leading makers of such cooling kits. Yet it is impossible to form a final opinion about a system without testing it in practice.

So we built a testbed around one of the hottest processors of today. It is the 3.8GHz model on the Prescott-2M core. The testbed was configured as follows:

• Intel Pentium 4 670 CPU (LGA775, 3.8GHz, 2MB L2 cache)
• ASUS P5WD2 Premium mainboard (LGA775, Intel 955X chipset)
• 1024MB DDR2-667 SDRAM (Corsair CM2X512A-5400UL, 2x512MB, 4-4-4-14)
• PowerColor RADEON X800 XT graphics card (PCI Express x16)
• Maxtor MaXLine III 250GB hard disk drive (Serial ATA-150)
• Microsoft Windows XP SP2

We performed our tests at the regular frequency of the CPU and in the overclocked mode. In the latter mode the frequency of the CPU was increased to 4.3GHz (Vcore was increased to 1.4V).

Several auxiliary programs were employed for the testing. First, we used the ASUS PC Probe II utility for reading the CPU temperature data. Then, although we disabled the thermal throttling feature in the mainboard’s BIOS Setup for the time of our tests, we still checked the real frequency of the CPU with the help of the ThrottleWatch program.

To heat the CPU up we used the S&M utility which is currently the most advanced tool for that purpose. We also ran the popular Super PI benchmark to create a high, but not extreme load on the processor.

Liquid-cooling systems require special test methods. For example, it takes quite an amount of time (about 2-3 hours in the given case) for the whole volume of liquid to get warmed up. So, the temperature data below reflect the CPU temperature as it was established after the system’s working for 3 hours in the specified mode.

Since the radiator of the COOL by Corsair system is supposed to be outside the system case, we can say that it transfers heat from the CPU to the outside. Thus, the type of the testbed (open or closed) isn’t important for testing such a system. We used an open testbed, so the air coolers taken for the sake of comparison had a certain bonus.

We compared the COOL system with the results of a popular air cooler from Zalman, the CNPS7700Cu model. We also tested the reference cooler Intel supplies for its Pentium 4 670.