In mid-2015, we published a comprehensive guide to CPU coolers, benchmarking six coolers featuring vastly different designs, from a low-profile 92mm model to a dual 120mm liquid cooler, using Intel's Haswell-based Core i7-4770K as the test platform. Among those six disparate coolers, we found one that truly stood out: the Thermalright Macho Rev. B. We followed up on that article with a comprehensive shootout among high-end coolers, starting at $50 and going up over $100, using the hexa-core Intel Core i7-5820K as the test platform. Among the coolers in that article, we selected the Noctua NH-U14S as the overall winner. In the meantime, we've also published a few one-off reviews looking at individual new cooler releases. But the truth is that our readers are always going to best served when we compare a number of products, and more importantly, when we provide a deep dive into why some coolers perform better than others in a given test scenario.
Our keen interest in CPU cooler performance dates back to a semester of college spent in a fluid dynamics lab, studying the ebbs and flows of gases and liquid over and through various surfaces. While we certainly can't at this point give our readers a primer on the physics of fluids, one thing is for sure: they don't always travel in a straight line. And that's what's so critically important to understand when looking at CPU coolers. If it were just a matter of pushing as much air as possible through as large a chunk of metal as possible, there wouldn't be much competition left in the CPU cooler market.
Here's why that just isn't the case: as you create ever-deeper fin arrays, you significantly increase drag, meaning that it becomes harder and harder to get air to travel all the way through the cooler and out the other side, which of course is the only way you can get additional air in to provide continued cooling. As they say, you can't get around the laws of physics, and this is why we've seen a huge proliferation of 140mm-based CPU coolers over the past few years. You see, building a wide fin array doesn't increase drag in the same way that building a deep fin array does. All things being equal, a tall, slender heatsink with a 140mm fan will always outperform a short, thick heatsink of the same mass cooled with a 120mm fan.
But of course there's even more to it than that, which is why we find ourselves here testing five 140mm CPU coolers, looking to get closer to unlocking the secret to perfect cooling. Since our last roundup, we built a new benchmarking platform, based on Intel's Skylake platform. And to put it simply, our Core i7-6700K runs much, much cooler than previous CPUs, changing the cooling equation substantially. Yes, you can go overboard with ultra-high-end coolers that could easily cool hotter-running CPUs from a few years ago, but why would you? Still, we decided to bring back our two previous champs, the midrange Macho Rev. B, and the high-end NH-U14S, and pit them against three of the latest entries in the 140mm cooler arena: the budget-priced Silverstone AR07, the midrange Reeven Ouranos, and the high-end Noctua NH-C14S.
OK, we know what you're thinking. Silverstone, sure, Noctua, of course, but who the heck is Reeven? Well, we were wondering the same thing when a Reeven rep came to us with the opportunity to check out one of its new products. Turns out the Reeven (or as the company styles it, REEVEN), is a new line of high-performance PC cooling products spun off of a long-time player in the PC market, Scythe. And Scythe is a name we know very well, so we accepted the offer to review Reeven's latest, even if Reeven is just barely entering the U.S. market at this point.
So, to summarize, here are the five coolers we're testing, along with their retail prices as of our publication date:
- SilverStone AR07 - $40 (special thanks to SilverStone for providing this review sample)
- Reeven Ouranos - $49 (special thanks to Reeven for providing this review sample)
- Thermalright Macho Rev. B - $53 (purchased at retail)
- Noctua NH-U14S - $70 (special thanks to Noctua for providing this review sample)
- Noctua NH-C14S - $80 (special thanks to Noctua for providing this review sample)
As we noted above, we have a new benchmarking system using Intel's Skylake platform, and as we quickly discovered, there really is a difference between generations when it comes to cooling requirements. That's why other review sites that keep their platform consistent (e.g. using a Core 2 Duo for the past ten years) in order to compare new products to older ones become much less relevant over time. When a new platform comes out, we'll always re-test our previous top picks along with new models on it, rather than stick with an obsolete platform. Yes, that means we don't have the dozens and dozens of models that you'll see listed in reviews on some other sites, but frankly, we think loading up a chart with a lot of old models that are no longer sold or no longer competitive just doesn't make sense.
Here's the system we used to rate our contenders:
- CPU: Intel Core i7-6700K (overclocked to 4.4GHz)
- Motherboard: Gigabyte GA-Z170X-Gaming 6
- Memory: Corsair Vengeance LPX 4x8GB DDR4-3000
- Solid-State Drive #1: Samsung 850 Evo M.2 500GB
- Solid-State Drive #2: Crucial MX200 1TB
- Video Card: EVGA GeForce GTX 970 FTW 4GB
- Power Supply: EVGA Supernova 850 GS
- Case: Phanteks Enthoo Evolv ATX
- Operating System: Windows 10 Home
Note that we chose to provide complete benchmarks of our CPU only in an overclocked state, providing only a baseline figure running at stock speed using the least expensive cooler. We think it will become abundantly clear that all of the coolers we tested are overkill if you're not overclocking, and even more so if you're using one of Intel's clock-locked dual-core or low-wattage quad-core processor.
Another issue that's we've been grappling with is how to create a basis for comparison between various CPU cooler models that run their fans at different speeds. Most reviews simply run fans at maximum, show the results, and perhaps on the following page provide noise data. Frankly, this just isn't good enough. Performance data divorced from noise data is meaningless, and it has encouraged manufacturers to "juice" the benchmarks by shipping coolers with ever-faster fans. Fast fans may increase overall performance, but they come with diminishing returns (again due to the laws of physics!), and more importantly, deliver a serious penalty in terms of noise. Therefore we ran our coolers through two separate sets of tests. First, we recorded data using motherboard PWM fan controls, which is the default for most systems, and therefore likely to be the setting most commonly used by PC builders. Second, we ran all of our coolers with their fans set to 1000RPM, which we view as a reasonable speed for extended use. In so doing, we are putting the coolers on equal footing (assuming their fans are generally up to the task), and truly shedding light on the effectiveness of their thermodynamic design.
The other major change we've instituted for our air cooler benchmarking is to shut off all system fans during testing. Both our power supply and video card remained idle during all measurements (when our power supply did heat up enough to require its fan to spin, we stopped testing). And we went ahead and unplugged all five of our case fans. We wanted to give our readers an accurate sense of both the effectiveness and noise levels of our coolers. Having other fans running makes assessing noise levels very difficult. The one drawback to this approach is that the case interior slowly heats up during the course of testing. We therefore performed our tests in the same order for each of our coolers, going from idle to moderate load to maximum load.
We also want to make clear that we do not use temperature "corrections" which, in a word, are methodologically inaccurate. Yes, it would be easy for us to just report the delta from ambient, as many sites do, or worse, factor in adjustments as the ambient changes, but basic thermodynamic principles suggest that using either shortcut method would void the results. A 2°C difference in ambient has a varying effect as a CPU's operating temperature changes, making any conversion worthless from our point of view. It may well increase idle temperature readings by 2°C, while it will have little to no effect on load temperatures. Our ambient was 68°F throughout our tests, and when the room temperature changed (as it in fact does as heat is produced during the tests), we stopped until it returned to 68°F +/- 0.5°F.
Why are we going into so much detail regarding our test methods? Because benchmarking CPU coolers is quite different from benchmarking the speed of components like CPUs and GPUs, despite the fact that most tech websites assess them similiarly. We felt a method taking into account the mechanical engineering aspects of CPU coolers should be applied. In other words, we want to do more than throw up some numbers, we want to provide insights into why the numbers are what they are, based on the differing design approaches used by each cooler manufacturer.
All right, with that introduction out of the way, it's time to get into the nitty-gritty of our results, starting with a review of installation!