When it comes to cooling performance, I don’t just test on one CPU. I run three different load types both at 100% fan speed and with the stock fan profile our motherboard provides. I do this across three different CPUs to get an idea of what to expect depending on what you are using the Pure Loop on. This includes the 3900X, 3600X, and then the 3400G. I start with the AIDA64 Stress Test using the FPU load which is the most demanding. Because Ryzen CPUs like to cap at a specific temperature and downclock for these extra demanding workloads I also keep track of the clock speeds. For the ML280 Mirror, I was especially curious how it would compare with the Gamer Strom Castle 280EX which has the same size radiator and is a similar design overall. My first tests were in AIDA64 with its Stress Test using the FPU or floating-point unit workload which heats things up the most but is more of a worst-case situation not what you would see in a lot of PC activity unless you are rendering. With the stock fan profile, the ML280 matched the 280EX on the 3900X and came in one degree higher with the 3600X and 3400G. With the fans and pumps turned up to 100% helped the ML280 edge ahead on the 3900X and matched on the other two CPUs. As for clock speeds, the 3900X didn’t boost as high, which most likely accounts for the 1 degree of temperature difference. But on the 3600X the ML280 matched and with the 3400G it came out a high higher on clock speed.
For my next test, I wanted a more realistic, but still extremely demanding workload and Blender fits the bill perfectly. With the stock fan profile, the ML280 didn’t perform as well as the Castle 280EX except with the 3400G. Cranking the fans up helped and it matched the Castle with the 3900X but the 3600X still ran warmer as did the 3400G. Like with the FPU testing, with the 3900X it clocked a stop lower, matched on the 3600X, then had a better clock speed on the 3400G.
For my last round of testing, I wanted something a little more tame and more realistic to the actual in game or everyday use that most people will see. For this I used AIDA64’s Stress Test again, but with the CPU workload. This thankfully doesn’t need clock speeds tracked so it is a lot easier to read as well. So with stock fan speeds on the 3900X, all of the AIOs have came in at 61c. Turning the fans and pump up on that CPU ended up having the two 280mm coolers at 58c, the older 240mm at 59c, and the new Pure Loop edged out another degree over the larger coolers here. With the 3600X the ML280 performed best with the stock fan profile and even edged out a degree at 100% fan speed when the three others all matched. Then the 3400G the ML280 was two degrees higher than the Castle 280EX with the stock fan profile but matched it at 100% fan speed.