We always use Gelid GC Extreme or Noctua NT-H1 thermal paste (almost identical performance) to make sure testing reveals the efficiency of the tested coolers not the efficiency of the bundled thermal paste.
Prime 95 is run for 10 minutes and then the average maximum temperatures as recorded by CPUID HWMonitor are noted
The system is left in idle for 10 minutes on the desktop and then the average minimum temperatures as recorded by CPUID HWMonitor are noted
Fans are always left to operate at PWM speeds, if this is not supported then 100% fan speed is used
Delta temperatures are always used (Observed temperature minus ambient temperature) and we keep the ambient at 22 (+/- 1) degrees for all testing although Delta temperatures correct for the changing ambient anyway.
The average temperature across the four cores is taken on our quad core processor
Acoustic measurements are taken 10cm away from the CPU cooler with the VGA fan disabled
The cooling performance tests are run at stock 3.4GHz (with Intel Turbo up to 3.8GHz) and overclocked 4.5GHz settings
All other coolers in the graphs have been tested under identical settings
Each test is repeated 3 times for consistency of results
There is approximately a 1 degree celsius margin of error in our temperature recording software
There is approximately a 1dBA margin of error with our decibel meter