Performance Testing

When I get complete kits or pre-filled AIOs such as this one, I typically test every individually moving component in addition to the entire cooler as a whole. However, a few things precluded me from doing so here, not least of which is that the unit I have here does not use the final version of the pump Alphacool went with. This still has the newer DC-LT 2 pump based on their older DC-LT pump I was moderately pleased with. The newer versions do seem to be better in both pure noise and also from a performance/noise basis, and the DC-LT 2 at 1600 RPM at least is pretty quiet. The retail pumps are rated at 2300 RPM (+/- 10%), so while there will be some more performance (think 0.1-0.3 °C on average), I am unable to say whether the pump at 2300 RPM will still be quieter than the fans.

The fans, on the other hand, were not tested because of time constraints, since the product arrived a couple of days before I was due to leave the country for what turned out to be a very long trip. We have previously tested these fans on the Alphacool Eisbaer Aurora, which uses the same ST30 platform for radiators, and I refer you to this page for more about the fans. It's obvious by now that the fans are the louder of the moving parts, even with the two fans here relative to the three there.

GPU block testing was also impossible individually, but mostly because I had a specific GPU here that Alphacool was kind enough to provide, since they were using this same model and GPU and did not yet have the reference PCB versions for NVIDIA and AMD cards. So with no other block here for the ASRock Radeon RX 5700XT Taichi, I skipped straight to cooler testing. This was done similar to how I test GPU blocks normally, as outlined here, except with a recorded temperature delta relative to ambient instead of the coolant, with the onboard thermal sensors coupled with our own GPU-Z instead of my own aftermarket ones, and the regular BIOS (not the quiet BIOS). The pump was set to 100% on the Eiswolf 2 since it was fairly inaudible over the fans anyway, and the fans were set to ~72% PWM so as to be at nearly the same noise level as the stock cooler at this load. To do this, cooler noise was also measured at the same testing conditions separately, inside an anechoic chamber at 19 dBA with the probe held 6" away to maximize SNR. The stock cooler was determined to be at approximately 44.8 dBA, and thus the fans on the Eiswolf 2 were tuned down until the cooler noise was the same.


As mentioned above, these are temperature readings that are normalized for noise already. So with that in mind, we clearly see the Alphacool Eiswolf 2 performing better across the board compared to the stock cooler. This in itself should not be a surprise, and the margins themselves are fairly impressive too. Interestingly, the relative gain going from air to water here is more with the GPU than the VRAM or VRMs, meaning Alphacool could do better optimizing for the latter. But it is not practical, so there is little reason to do so. Both coolers have more performance left in the bank since both sets of fans can be manually set to 100%, but I suspect many will just go with a fan curve to suit their needs anyway, and prioritize noise when thermal throttling is not as important for GPU overclocking as power limitations tend to be these days.