We tested the thermal compounds using the same testbed system that we currently use to test CPU coolers, which is fully described below. Our Core i7-860 (quad-core, 2.8 GHz) CPU, which is a socket LGA1156 processor with a 95 W TDP (Thermal Design Power), was overclocked to 3.3 GHz (150 MHz base clock and 22x multiplier), and we kept the standard core voltage (Vcore). We used a Zalman CNPS9900 MAX CPU cooler. The only different part in each test was the thermal compound itself.
We measured temperature with the CPU under full load. In order to get 100% CPU usage in all threads, we ran Prime 95 25.11 (in this version, the software uses all available threads) with the “In-place Large FFTs” option. For each test, we applyied the same quantity of thermal compound (about the size of a grain of rice) at the center of the CPU, as shown in Figure 12.
click to enlarge
Figure 12: Applying thermal compound
After each test, we checked the base of the cooler, making sure the quantity of thermal compound was optimal. The thermal compound must be spread evenly on the metallic part of the CPU, without exceeding it, creating a thin layer. The “fingerprint” shown in Figure 13 illustrates that the compound was properly applied.
click to enlarge
Figure 13: CPU “fingerprint,” showing the thermal compound was correctly applied
Room temperature measurements were taken with a digital thermometer. The core temperature was read with the SpeedFan program (available from the CPU thermal sensors), using an arithmetic average of the core temperature readings. During the tests, the left panel of the case was open.
We also tested the system with no thermal compound on the CPU.
Operating System Configuration
- Windows 7 Home Premium 64 bit
Since both room temperature and core temperature readings have 1 oC resolution, we adopted a 2 oC error margin, meaning temperature differences below 2 oC are considered irrelevant.