The Overclocked Benchmarks
When push comes to shove, what really matters to enthusiasts is cooler performance with an overclocked CPU, and that’s what you’ll see on this page. We took our Core i7-4770K to 4.4GHz on all four cores, with Hyperthreading enabled, and ran it through our battery of tests. Note that our particular CPU sample was able to achieve this overclock at a relatively-low voltage of 1.18V, just 0.07V higher than stock. That being said, the combination of the higher clock speed and the exponential increase in temperature with voltage will spell doom for a few of the coolers in our round-up, and it’s not just the ones you would expect.
To be clear, if you’re overclocking, you could jeopardize your system’s lifespan if you don't carefully monitor temperatures, so please be aware of the risks. The effectiveness of your CPU cooler becomes much more critical when overclocking, and for the most part, we do not recommend overclocking when using the stock Intel cooler, although as you'll see, it can be done under certain conditions.
Above you can see the idle measurements running an overclocked CPU (we’ve skipped the maximum RPM measurements on this page, as they really aren’t relevant at idle). If you’re using “adaptive” voltage control for your Intel CPU overclock, the CPU shouldn’t be running at much higher voltages at idle versus a stock setting. And even if you have a fixed voltage, idle state still won’t be where your system will be at risk, although fans could run at slightly higher RPM. Because we only tested with adaptive voltage, nothing changed significantly between our stock and overclocked idle state numbers. Thus, the results largely mirror what you saw on the last page. As before, the two best coolers in regard to idle noise, other than the Intel cooler, are the Noctua NH-U12S and the Thermalright Macho. But what will separate winners from losers for extreme overclockers are the load numbers, and that’s what you’ll see below, starting with the PWM-controlled results.
OK, we’re seeing some failures here in Intel Burn Test. The two low-profile coolers, from Intel and Noctua, failed within seconds, as the CPU exceeded 100C and throttled. That’s not a problem from our point of view, as these coolers were not designed (nor should they be utilized) for extreme overclocking. More surprising was that a couple of our Noctua coolers met their limits, with the mighty NH-D15 in single-fan vertical mode failing near the end of our Intel Burn Test run, despite hitting a maximum fan speed of 1505RPM.
When pushed to the extreme, it’s clear that the potential of the NH-D15’s massive heatsink is wasted when oriented vertically, and that even much smaller heatsinks work better in a horizontal position. The basic reason for this is that when mounted horizontally, a heatsink works in tandem with the natural front-to-back airflow of modern cases, which often feature one or two intake fans plus an exhaust fan. Even when cases have a top-mounted exhaust fan, as many new cases do, the primary airflow is still from front to back, and always will be until cases start featuring multiple bottom-mounted fans (one of the few examples of such a case is the Silverstone Raven, which also turns the motherboard 90 degrees). Another problem is that when mounted vertically, the heatsink is drawing air in from the single hottest point in the entire case: the back of the video card. While this is the only heatsink we tested in a vertical orientation, we're pretty confident based on our results that all tower heatsinks will be compromised when used in such a fashion.
While the vertical NH-D15 is a disappointment, there are some real standouts among the bunch here. We have to call out the Corsair H100i for its excellent performance-to-noise ratio in an overclocked situation. It’s among the quietest coolers in this test, despite garnering the best results overall. Pity, then, that it can’t shut down its pump at idle, because it’s one of the loudest coolers when not under stress. Also impressive is the Macho, which comes through again with some very impressive low-RPM numbers. It’s able to match or beat the NH-U12S’s results while never exceeding the NH-U12S’s idle-state noise level. That speaks to the power of a 140mm fan and a massive (if somewhat cumbersome) heatsink. Finally, we’d skip right past the dual-fan NH-D15 numbers and look at the same cooler with a single fan. It’s actually quite impressive, nearly matching the H100i in overall effectiveness (and its Battlefield 4 temperature is only higher because it was running at very low RPM in that test). For massive cooling potential under adaptive PWM control, it’s a nearly perfect solution. Frankly, we wish Noctua would sell the NH-D15 at a reduced price with just a single fan. In comparison to the dual-fan setup, it would be easier to install, so much quieter, and for most users, plenty capable. Imagine an $80 single-fan NH-D15, and you'll be imagining air cooling Nirvana.
Our final test takes runs the fans on all of coolers at maximum RPM. Keep in mind that the two larger Noctua coolers will have similar results here as compared to the PWM benchmarks because they were able to spin all the way up to maximum RPM even under motherboard control. That is a key feature in an overclocked situation. Indeed, even at their fixed maximum RPM, the Noctua NH-U12S and NH-D15 only end up slightly cooler, owing to the fact that at maximum RPM, they are actively suppressing heat all the time, rather than just reacting once the motherboard detects a high-heat situation and sends them more voltage. On a separate note, as much as we like the Macho for its low-RPM performance, we just can’t get behind its high-RPM performance. It loses its low-noise advantage, while barely out-performing the more elegant NH-U12S.
By the way, in this testing situation, we’re going to call out the NH-D15 for an honorable mention. In single-fan mode, it proves that it's indeed stronger overall than the Macho, at least when not in a vertical orientation, and when the second fan is strapped on, it performs nearly as well as the H100i here. Of course, it's very loud, but it doesn't create quite the same deafening racket as the H100i in software-controlled “Performance” mode. Faced with a lot of voltage, and therefore heat, the H100i automatically ramps its fans up to extremely high levels (remember, these are 2700RPM-rated fans), exceeding 50dB, the one and only example of this happening in our testing. Quite frankly, it’s unacceptable, and it’s no surprise that Corsair has outfitted the newer H100i GTX with lower-RPM fans, rated at a maximum of 2435RPM. That being said, we were quite pleased with the H100i in “Quiet” mode, and would take it over the dual-fan NH-D15.