To test all CPU coolers, we devised an easily repeatable test with no variables other than the coolers themselves. This ensures that figures from every cooler we test are comparable with each other.
Using an open-air test bench, we deploy an Intel Core i7-4790K plugged into a Gigabyte Z97X-SOC Force motherboard. Alongside this is 16GB of 2400MHz Corsair Vengeance DDR3, as well as a 120GB OCZ Trion 150 SSD. Powering everything is a Corsair AX1500i PSU.
The test process
Testing coolers involves taking a total of 4 temperature readings per cooler. First, we measure the idle temperature of the i7-4790K at stock speeds (turbo boost disabled), before measuring its temperature under load at stock speeds. Next, we overclock the CPU to 4.5GHz using a 1.3 Vcore, ensuring greater heat output. In its overclocked state we then measure the idle and load temperatures of the CPU again. The figures we present are temperature deltas – meaning we take each temperature reading and minus the ambient temperature from it. This allows us to test in an environment that is not temperature-controlled.
To ward off potential comments or questions, we know 4.5GHz using a 1.3 Vcore is not the ‘best’ overclock – this particular CPU could reach that frequency at closer to 1.25 on the Vcore, which is more efficient. That is not the point, however. We are trying to stress the coolers to see how they deal with excess heat … hence the higher than necessary Vcore.
Where possible, each cooler’s fans are plugged directly into the motherboard using the CPU_Fan or CPU_Opt headers. Some AIOs, however, ship with their own fan controllers or PWM hubs. If we are unable to plug the fans directly into the motherboard, it is specified in the performance section of the review.
An idle reading comes from leaving Windows on the desktop for 15 minutes. A load reading comes from running Prime95’s (version 26.6) Small FFTs test for 15 minutes – enough time for temperatures to plateau. We use HWInfo64 v5.82 to measure temperatures.
We measure noise levels with our sound meter positioned 1 foot away from the test bench. The peak noise level is presented. We have measured the noise floor (i.e. ambient noise levels of the room) to be 34 dBa, and when tested at stock clocks, the vast majority of coolers do not emit noise levels above this figure. Thus, we present one chart with noise levels taken during our overclocked CPU test run.