Performance (Fixed 1000 RPM)
As we discussed in the introduction, we've long-searched for a CPU cooler test scenario that was methodologically sound. The most common test you'll see in tech publications is benchmarks using the maximum fan RPM, which is quite simply useless. Any of these fans when run at 100% will both be far more powerful than you need and far louder than you can tolerate. And it leads to an arms race by manufacturers to equip their coolers with faster fans than everyone else. Sadly enough, even Noctua, with its immense technical prowess in air cooling, isn't immune to this; it equips its coolers with non-retail versions of its 140mm fans that operate at 1500RPM instead of a more reasonable 1250RPM.
While the ideal test setup would probably be to pick a decibel level and find each coolers best results at whatever fan speed this equates to, dialing in a constant sound level for various models in various tests would be a nearly impossible task. So we've done the next best thing: we've found the fan percentage setting for each model that equates to 1000RPM, a rotational speed that we think represents the sweet spot for 140mm fans. They move plenty of air at this speed due to their large size, and when run faster, they become rough-sounding, just like their inferior 120mm cousins.
First up, idle numbers.
We'll admit that these results are mostly irrelevant; if you're going to bother dialing in a custom fan speed, you're most definitely going to run an ultra-low speed at idle. But that being said, we can see Noctua and Reeven pulling ahead here in terms of acoustics (which will remain the same in the graphs that follow, given that the RPM is fixed), while the Thermalright Macho again wins in terms of temperature due to its large size.
But take note: in actual use, we found the Noctua NH-C14S to sound louder than our test results would suggest. How is that possible? Well, due to its different orientation, fan noise is projected in different directions. It does not project much toward the front of the case, but certainly projects just as much upwards (and in our setup, that meant towards our ears), and so we found that despite a very low decibel reading, it wasn't actually quieter than the NH-U14S. That shouldn't be surprising, given that they use nearly identical fans.
Below, we have what we believe best represents real-world loads, CPU-Z.
The relatively close results here serve to illustrate the importance of considering fan RPMs. Despite vastly different designs, sizes, and prices, these heatsinks are all operating within a very narrow range, and are all more than adequate for the job at hand. But when you look at noise levels, certain results stand out as better than the rest, namely the three in the middle of the graph (and the price spectrum).
Finally, we illustrate cooling performance faced with the the artifically-high load generated by Intel Burn Test.
Here we see the Noctua NH-U14S providing very low temperatures and very low noise. This, dear readers, is what makes the NH-U14S so special. If you look back at the previous page, you'll actually see it was running both hotter and at higher noise levels. An error on our part? No, in fact, it has to do with the dynamics of PWM testing. The NH-U14S ramps up and down like crazy while using PWM controls, getting close to its maximum RPM, but doesn't react quite fast enough to keep peak temperatures at bay. At a fixed 1000RPM fan speed, the NH-U14s is cool, calm, and collected, achieving equilibrium with its target, waste heat. By the way, the Reeven Ouranos is deserving of special mention here for its exceptionally-low noise reading and very good temperature result, while the Thermalright Macho gets the nod for bang-for-the-buck performance at slightly higher noise levels.
While we haven't wanted to rub it in throughout our tests, we'll just mention that the SilverStone AR07 is clearly outmatched in this roundup, and while it is the least expensive entry, several models that cost just slightly more provide much, much better performance.