Test protocolWe took two separate readings to evaluate the various cooling systems covered in this report.
First of all we took temperature readings to measure cooling capacity. For this we used a Core i7-2600K CPU loaded with Prime95. The temperatures at idle weren’t taken as they aren’t particularly representative given the low consumption levels of such a processor at idle: pretty much any cooler can dissipate ten or so watts silently. This CPU was used in three configurations:
- At the base clock with energy consumption measured at 77 watts by the ATX12V
- Overclocked to 4.5 GHz and 1.35V, with energy consumption of 134 watts
- Overclocked to 4.8 GHz and 1.45V, with energy consumption of 175 watts
All the fans were tested with DC at four different voltages: 5V, 7V, 9V and 12V. Four gradings are given:
- N/A: The fan isn’t working at this voltage
- Failure: Cooling insufficient, the CPU throttles power
- Temperature in red: Above 75℃ (temperature obtained in load on an i7-2600K with a box cooler)
- Temperature in black: Temperature reading given
A few explanations follow: During a failure, the cooler doesn’t manage to maintain the temperature of at least one of the cores under 95°C. The CPU then lowers its clock in an attempt to maintain temperatures at a level that won’t be damaging.
The temperature reading given corresponds to the average temperature of the four cores as reported by HWMonitor. The tests were not carried out in a casing. The room temperature was 22.5°C. For a temperature reading to be considered as acceptable, we leave ourselves a leeway of around 20°C, which corresponds to the temperature obtained using the Intel box cooler at 12V on a 2600K at its base clock: 75℃.
Noise levels were measured with a Cirrus Optimus CR152A Class 2 sonometer, which can measure levels as low as 21 dBA, the lowest level permitted by the room used for testing. This is the second time we’ve used this high end sonometer (the first being during our roundup of entry level coolers
). Up until now we were using more standard sonometers that allowed us to measure down to levels of between 30 and 35 dBA. We’re planning to roll out usage of this type of sonometer in our other tests so as to be able to publish results that are comparable from one domain tested to another.
The cooler was placed on the ground with the sonometer 50 cm away and raised up 25 cm to take the reading. The solutions measured at between 21 and 22 dBA can be considered as quiet. Up to 25 dBA and the cooling is very discreet. Between 25 and 30 dBA can be considered to be discreet. Between 30 and 35 dBA can be considered standard and between 35 and 40 dBA is starting to get noisy. Beyond 40 dBA can be considered as difficult to put up with for a computer, though this is of course subjective and depends on several factors such as the regularity of the noise or the environment.