First up, the fan itself. Testing was done with the fan mounted on the AR08 heatsink in a push configuration, and controlled by an Aquacomputer Aquaero 6 XT in PWM mode which also enabled RPM readout. After this, linear airflow was measured using an Extech 45158 Thermo-Anemometer 6″ away from the fan such that it measured the airflow in feet per minute through the heatsink. Fan noise was measured in an anechoic chamber of size 5′ x 8′ with ambient noise level ~19 dBA and a sound probe held 6″ away to measure the sound volume in dBA accordingly with nothing else powered on so the fans were the only source of noise.
Silverstone has always made decent fans and this is no exception with a mostly linear PWM-RPM response curve, albeit with a smaller duty cycle range than I would like. The min and max RPMs match the rated values very well too, well within the +/- 10% standard deviation I allow. As I suspected, at full speed this is audible for sure but on par with larger fans running at similar or higher speeds even.
Since this question will inevitably arise, this is how the AR02 fan did on the AR02 heatsink:
Very, very similar. The only real difference is a smaller min RPM here although you really would not want to run this fan at minimum RPM unless you are just testing it out- you need the higher speeds to effectively cool most CPUs.
On to the CPUs then. As mentioned before, I will only be testing the i7 4770k and the Pentium G3258 here as I don’t think anyone will realistically use this to cool an i7 5960x at ~4.3 GHz. So here are the tests done:
4 core: Intel i7 4770k (4.5 GHz at 1.25 Vcore) with dual channel DDR3 memory
(a) Fan at 50% (Test 1, cooler noise 29 dBA)
(b) Fan at 100% (Test 2, cooler noise 46 dBA)
2 core: Intel Pentium G3258 (4.5 Ghz at 1.2 Vcore) with dual channel DDR3 memory
(a) Fan at 50% (Test 3, cooler noise 29 dBA)
(b) Fan at 100% (Test 4, cooler noise 46 dBA)
Each test was done in a climate controlled hotbox at 25 ºC ambient but the results will still be reported in terms of a delta T (component – ambient) in ºC. The ambient temperature should be nearly the same throughout the box with a +/- 0.05 ºC variability at most but still the ambient temperature sensor was held in front of the intake fan(s) as should be the case with CPU coolers in my opinion. If including the ambient temps into play, you are now introducing another variable into play here- the case/test bench air cooling layout- and this has no role in the review of a cooler. Each test/scenario was repeated three times and the average value taken to account of possible issues with mounting and TIM. Speaking of TIM, the stock TIM provided is not enough for installation demo + these 9 tests and so I used my reference TIM (Gelid GC Extreme) for the tests with cure time accounted for by measuring values after steady state was achieved post burning in with an Intel XTU run. This may well be a variable that benefits the AR08 as the stock TIM on a sample test performed ~2 °C worse so keep this in mind. Aida64 was used to measure the core temperatures.
Not too bad, but this only tells half the story as we need context to complete it. So let’s see where this stacked up against other coolers in terms of both cooling and noise. Results from Test 2 here were used for comparison. Please note that lower is better here.
The Silverstone AR08 is one of the weaker performers thermally, but this was not a surprise considering it is the smallest cooler tested so far. The older AR02 is just behind it, so yes there is some improvement to be seen. The cooler is also not as noisy when context is provided, and comes in the middle of the pack. So there is opportunity to use an even faster fan and get better thermal performance at the expense of added noise or, more likely, pair this with a low TDP processor and keep the fan at lower fan speeds. You will lose ~2-3 °C but at the same time also lose ~15-20 dBA of noise from the cooler which is a massive amount. Use the cooler as common sense dictates and it will work great. Try to push its limits and you might run into trouble.