The Stock Benchmarks
Now we get into the meat of the article – the performance benchmarks. That being said, we’re going to start with some pretty mundane charts – the idle performance numbers at stock settings. It’s important to understand that looking at idle numbers to pick out winners can be misleading – most heatsinks perform within a narrow performance band at low temperatures, and furthermore, many models are optimized for high temperature situations, and will often underperform simpler coolers at low temperatures. The most important thing to look at, then, is the noise levels. Frankly, it’s the only thing that matters when it comes to idle-state performance.
Each of our charts shows both temperature and sound level data. We know that makes them a little hard to read, but to be honest, heatsink performance must always be judged both by the ability to control heat and the ability to limit noise. Too many reviews separate out this data, or worse yet, don't include noise data, and that just encourages manufacturers to ship their products with increasingly fast (and noisy) fans. We're hoping that by addressing both factors simultaneously, overall heatsink performance will be clearer.
Right off the bat, we see that all of these coolers work just fine at idle, even when controlled by the motherboard (or in the case of the Corsair H100i, when set in low-RPM “quiet” mode via the CorsairLink software). But the noise levels are not all the same. As you’ll see, we’ve included RPM ranges for all of the coolers that are motherboard controlled, and at idle, you should assume that they are all at their lowest possible speed. But that speed differs dramatically from cooler to cooler, and so you have the 727RPM Thermalright Macho idling at 37.5dB (our test system’s noise floor), while the dual-fan Noctua idles at 1038RPM and registers 41dB, proving to be even louder than the Corsair Hydro H100i, which is at a disadvantage here because its pump always runs at full speed (the fans are actually spinning at around 1000RPM in quiet mode). Also keep an eye on the NH-D15 in vertical single-fan mode (with its fan blowing up towards the top of the case). It’s the loudest of the single-fan models, and also performs the worst. Not a good start, and unfortunately, it’s a sign of things to come. But Noctua does have one important piece of the puzzle nailed down, and that is fan quality. For example, the bearings in the Thermalright Macho's TY-147A fan are clearly of a lower grade than those in Noctua’s fans, making a slight scratching noise despite incredibly-low decibel levels. The human ear is more sensitive than a decibel meter when it comes to changes in pitch, rather than sound intensity, so while the Thermalright’s sound level is lower, it’s not necessarily better than the Noctua NH-U12S, for example, which is running over 200RPM faster at idle.
Unfortunately (or perhaps fortunately, depending on your perspective), not all fans are controlled in the same way by our motherboard’s PWM fan headers. Some of the fans could be pushed up to maximum RPM (the Noctua NH-U12S and Noctua NH-D15), while the majority would use a much narrower range, even when pushed to the extreme (that includes the Intel, the Noctua NH-L9x65, the Macho, and the Corsair). That’s great for limiting noise, but it does curtail heatsink effectiveness somewhat. So we also include charts for each of our tests running fans at maximum RPM. You can see those numbers below, which for idling with a stock-clocked processor causes a simply outlandish amount of noise. The Intel actually surprises with a fairly low maximum noise level of 41dB, which is actually equal to the NH-D15’s idle level. Yikes! So much for reducing noise with a high-end cooler. The rest of the single-fan coolers are all between 43dB and 44dB, although interestingly, at very different RPMs. The Macho’s big 140mm fan unfortunately is at a disadvantage in this test – it just isn’t very good at high RPM. Perhaps Thermalright intentionally designed it to stick to a narrow RPM band when under PWM control, because at around 1300RPM, it’s kind of a stinker, making more noise than Noctua’s smaller models at higher RPMs, and making as much noise as the Noctua NH-D15 single-fan setup despite a 200RPM deficit.
The biggest losers when running at maximum RPM are the ridiculously loud Corsair H100i and dual-fan Noctua NH-D15. Keep this in mind when you see the load numbers that follow, because from our point of view, running these coolers at maximum RPM full time would be an unacceptable solution.
For our load tests, we used two different benchmarks: a 20-minute run of Battlefield 4’s multiplayer mode to simulate a typical CPU-intensive application, and 20 loops (about 3 minutes) of Intel Burn Test. Intel Burn Test taxes all cores at 100% and adds an AVX2 component to simulate a maximum load scenario, and quite frankly would be impossible to replicate in any typical application. It’s the worst of all worst cases. Our temperature numbers and decibel figures are maximums measured during our tests, but note that we don’t provide noise levels for our Battlefield 4 run, as it would be impossible to distinguish between CPU cooler noise and the additional power supply and video card noise inherent to these runs. For most of the coolers, temperatures were low enough during a gaming session to keep PWM-controlled RPMs near their minimums (i.e., close to idle noise levels).
Unlike with the idle numbers, we do want you to look at temperatures here, because as you’ll see, they vary quite a bit. All of the coolers except the stock Intel cooler were able to get through this first round of load tests without problems, and even the Intel cooler was fine in a typical load scenario. We’d also like to highlight one standout among the coolers here: the Thermalright Macho. As we mentioned above, its fan stays within a very narrow range under PWM control, so it’s a good example of what a low-RPM 140mm cooler can do. In a word, it’s exemplary, achieving results between Noctua’s mid-sized and over-sized tower coolers, while besting them both in noise (and under-cutting them in price!). Another thing to notice is that once you move beyond the two compact coolers, performance tends to fall within a narrow range. It will be interesting to see if that remains true when we push into overclocked territory on the next page.
Here’s the last round of tests running our Intel Core i7-4770K at stock settings, this time with the fans on our coolers maxed out. This gives you a good sense of the ultimate cooling potential of each model, using one of the hottest running consumer-level CPUs Intel has released in recent memory. This time every cooler passes all of our tests, including the Intel cooler. Take note of the performance of the Noctua NH-D15 in single-fan mode. When mounted in a vertical orientation, it’s just not particularly good, losing to the smaller Noctua NH-U12S in both temperature tests while making the same amount of noise. And in the standard configuration, the NH-D15 is oddly a bit hotter in Battlefield 4 than it was with PWM controls, which can only be explained by slight test-to-test variation in CPU loads inherent in real-world (non-synthetic) testing. At the very top, the dual-fan NH-D15 and H100i trade blows for supremacy – we’ll call the Noctua the winner here for running a bit quieter, but overall, both are just too loud.
For PC enthusiasts who are more interested in an all-around balance of performance and noise levels for running a stock-clocked CPU, the results on this page are all you need to see. If we were to name a winner, it would be the Thermalright Macho, followed very, very closely by the Noctua NH-U12S, which is much easier to install, but with cooling potential that isn’t quite as high. If you’re running a stock CPU, the two high-end options (the NH-D15 and H100i), both of which run around $100, simply should not be considered. Spend the extra $40-$50 they’ll cost you over the NH-U12S or Macho on a better CPU or video card.