Blog

Jun 02, 2023

Noctua NH

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More. Table of Contents The biggest new heatsink from Noctua is the NH-D15. It is

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Table of Contents

The biggest new heatsink from Noctua is the NH-D15. It is even wider than the NH-D14, and it comes with two NF-D15 PWM fans that spin up to 1500 RPM. The NH-D15 also comes with two RC-7 Low Noise Adapters so that the heatsink can run silently. Is this the best air-cooled heatsink? Let’s find out.

The tandem tower approach to heatsinks dates at least to the LGA775 Thermalright IFX-14 of 2007. The twin tower Prolimatech Megahalems was released in early 2009, and was a big hit in its day.

The Noctua NH-D14 tandem tower heatsink came out in late 2009. As various people tested it and published their results, the NH-D14 began accruing massive acclaim, which has hardly died down to this day. It has the reputation of being a better cooler as the temps get hotter. The NH-D14 has become the gold standard for heatsinks, and rightly so – it is a fine heatsink. You don’t know how good a premium heatsink really is until you have compared it to a NH-14. We shall use a Noctua NH-D14 SE2011 as the comparator heatsink in this study.

The heatsink we are reviewing today, the NH-D15, is Noctua’s improvement over its previous top heatsink. Noctua says this:

Compared to the NH-D14, the NH-D15’s fin-stack has been widened from 140 to 150mm and its heatpipes are now spaced further apart, which allows for a more uniform heat distribution over a larger surface area and thus contributes to the NH-D15’s superior efficiency. Thanks to its recessed lower fins, the NH-D15 provides 64mm clearance for tall memory heatsinks in single fan mode, making it compatible with most high end RAM modules on the market.

The NF-A15 premium fans supplied with the NH-D15 support PWM for convenient automatic speed control through the mainboard. In addition, the maximum fan speed can be reduced from 1500 to 1200rpm using the supplied Low-Noise Adapter for even quieter operation.

Noctua say this about themselves:

Designed in Austria, Noctua’s premium cooling components are internationally renowned for their superb quietness, exceptional performance and thoroughgoing quality. Having received more than 3000 awards and recommendations from leading hardware websites and magazines, Noctua’s fans and heatsinks are chosen by more than a hundred thousand satisfied customers around the globe.

Note that the retail version of the NF-A15 is a 1200 RPM fan. It is still PWM, but slower than the fans that come with the NH-D15.

The NH-D15 comes in a nice box. This one still bears some postal service marks; but when they come direct from Noctua, the retail packages come inside another box, so you won’t have those marks.

Just to give you a sense of the size of the box, here is a picture of the NH-D14 and NH-D15 boxes next to each other.

Open the box and you will find three shallow boxes that contain the SecuFirm2 mounting systems (Intel and AMD), and the box of accessories.

Inside the box, the layer of shallow boxes sits on a flexible foam layer, which in turn sit on boxes containing the extra fan and the heatsink, which in turn sit on another layer of flexible foam.

Inside the accessories box we find a bag of four silicon vibration isolators, a pair of RC-7 PWM Low Noise Adapters, a tube of NT-H1 thermal interface material, a Y-cable, an L-shaped Phillips #2 screwdriver, and a pair of fan clips for the second fan. Inside the box for the SecuFirm2 mount we can see that two of the pieces are fixed in cardboard. The X-mount (you will see this later) is also fixed in cardboard with the smaller pieces in a bag. Under the fan clips are the folded Intel mounting instructions.

When you take out the box that holds the heatsink, you discover not only that the numbered flaps open at the top, but the side opens to reveal support for the recess in the fin stack. Note that to provide such support, the main NH-D15 body ships in this box with the center fan clipped into place while the second fan ships in its own cardboard wrap.

Let us build a SecuFirm mount and put a heatsink on it. We will start with an integrated backplate assembly – that is a backplate with integrally mounted bolts. Those tall bolts go through your motherboard. Then we put spacers on the bolts.

The backplate goes under the motherboard. The bolts go through the motherboard. The spacers go on top of the motherboard.

Then we put on the brackets. After that we put on the thumbscrews. It is all covered in Noctua’s installation manual (PDF here).

Then you put down your TIM, strip the plastic film off the NH-D15’s contact plate, and mount your heatsink. Fasten the two spring-tensioned mounting screws, tighten them down to their stops, and your heatsink is ready for its fans. Easy.

In fact, this is the easiest mount I have used.

The Noctua NH-D15 comes with a single fan mounted in the center. Later, we shall see how this arrangement fares, but for now you should at least see how it looks. In this picture, the air would go down.

Now to compare the NH-D14 (on the top) with the NH-D15 (on the bottom). You can see that they are both huge, but that the NH-D15 is a little wider.

The first picture compares a mounted NH-D14 with the mounted NH-D15. You can see the integrated nylon fan vibration dampers built into the NH-D14. Those were removed in the NH-D15 when Noctua went to soft silicon corner vibration dampers integrated into the NF-A15 fans that come with the heatsink. On the NH-D15 you can see where the cutaway portions of the fin stack allow for high RAM heatsinks up to 64 mm tall – on both sides. Note the backplates under the two heatsinks. The SecuFirm backplate is clearly bulkier than that of the SecuFirm2.

Comparing the SecuFirm with the SecuFirm2, you can see that the SecuFirm on the left has bolt heads that are captured by the rim of the X-shaped backplate. While this works OK for a vertical motherboard, if your motherboard is horizontal you will have a bit of a time getting the bolts through. Try holding a pair of them with your fingers on one side while you attach spacers, brackets and thumbscrews with the other. The SecuFirm2 on the right makes the job ever so much easier. The through-bolts are press-fitted to the backplate. Nice.

Finally, the wing nut reflection on the heatsink contact surface shows that, mirror-like as it is, it is really a milled surface. On the right you can just make out the curved ends of the milling lines.

The first picture shows a stock NH-D14 SE2011, with a 120 mm NH-P12 PWM push fan in front, and a 140 mm NF-P14 PWM fan in the middle. Notice the RAM here is ultra-low profile RAM. One of the benefits of lower voltage RAM is the ability to make RAM sticks ever shorter. There is simply no reason aside from marketing to have tall RAM heatsinks. The fans are using a Y-cable, which is part of the NH-D15 kit. However, because this motherboard has two CPU fan headers, we will later dispense with the Y-cable.

The second picture shows one of our comparison setups. The 1200-RPM NF-P14 PWM that came stock with the NH-D14 SE2011 this time is in front, acting as a push fan. An NF-P14r redux-1500 PWM fan (this is the only type of NF-P14 with a round frame that is still sold) is in the center. The key point to note here is that the ultra-low profile RAM allows the front 140 mm fan to sit as low as the center 140 mm fan. That ultra-low profile RAM makes it possible to run two 140 mm fans instead of one. You have the potential either for quieter running or more effective cooling, or both.

I did not have a pair of retail 1200-RPM NF-A15 PWM fans to use for testing, so I used a pair of 1200-RPM NF-A14 FLX fans on each of the heatsinks. The NF-A14 and the NF-A15 use the same impeller – they differ in the shapes of their frames with the A15 being 10 mm wider than the A14. The first picture shows the two A14s on the NH-D14. The fan clips are Megahalems 120 mm fan clips, which work perfectly with 120 mm push fans on the NH-D14. With the NF-A14 fans, the clips did not fit in all four screw holes, yet found places to grab on. Again, note that the low profile RAM allowed using 140mm fans – even square-framed 140 mm fans – at both push and center positions.

Finally, here is the NH-D15 with its two 1500-RPM NF-A15 PWM fans. Note first of all the position of the digital heat sensor (thermometer) in front of the push fan. That is its proper position for the subsequent testing. Next, note that the 140 mm push fan again has room to sit on the NH-D15 without being excessively elevated. If you are not already convinced of the utility of low profile RAM, perhaps this picture will do it. The fans themselves have integrated cushions that press against the heatsink so they will not rattle. These fans in particular were plugged into the motherboard through RC-7 Low Noise Adapters.

Noctua fan clips have evolved. The NH-D14 clips are inserted into push-in rivets that poke through the screw holes in each fan’s flanges. The flanges are thinner than those of normal fans, so the push-in rivets that come with the NH-D14 will not push all the way through the flanges of normal fans, unless you modify the holes by countersinking them. You can also replace these push-in rivets with longer ones.

One problem with Noctua’s original fan clips is that they stick out to the side where they hook onto the fin stacks. If you have a motherboard with the main GPU card in the top slot – in mATX boards, for example — a user would do well to put non-conducting tape over the clip to avoid a short circuit in the back of a GPU card. And note the depth of the holding slot in the fin stack. We will get to that.

Worse than the clip handles sticking out, from a user perspective, was the tightness factor. Sometimes the fans were hard to get off. First, you would have to pull the clip parallel to the fin stack (#1 in the picture below) through that deep slot on the fin stack. Then you would pull the clip away from the fin stack (action #2 below). Back in the days when Noctua shipped a third set of clips and push-rivets to any owner who asked, I remember having to use pliers on occasion to pull a clip far enough to free it. It was hard work.

Then Noctua developed a different clip – the right one in the second picture. They began shipping a pair of the new clips to those who wanted a third set of clips. One nice thing was that handle part of the clip was angled back, to keep it out of the way of the forward-angled clip for the middle fan. They also began shipping this clip with other heatsinks — the NH-L12, for example. Finally, this clip is much easier to pull out of the slot in the fin stack – no more pliers.

Now we look at Noctua’s third iteration of its clips — the clips for the fans on the NH-D15. First of all, the clips wrap around the outside of the far flange before inserting into a screw hole. Next, the fin stack has two rows of lumps – one to provide the clip with an anchor for its grip, and one to hold the handle.

Seen from above, it is clear that this clip clings more closely to the side of the fin stack. In the second image you can see how these fan clips work. There is no more two-step process. Just pull the clip, as shown by the small yellow arrow (small because it doesn’t take much effort to pull the clip away from the fin stack). The large blue arrow shows where this clip’s main effort is exerted. Although this process may seem trivial, the design of these clips is the result of some really clever engineering.

OCCT changed its algorithm between version 3.1 and 4.0 so that what had been a dependable straight line in OCCT 3.1 became a jagged line with hiccups in 4.x. I was unable to get OCCT 3.1 to run at all in Windows 8.1, so I had to use the later, inferior version. Other software did not stress the system as hard, except for Linpack. Linpack, however, has its own methodological problems, and until I solve them I will use OCCT as a reasonably uniform stress software.

The 4.5GHz OC was chosen because, with an ambient temp of 22 °C, the hottest core temp with my Megahalems was in the low 90’s with Linpack. This seemed to be a reasonable overclock. YMMV.

The temperature of the intake air (ambient) was measured at a spot about two inches (5 cm) from the center of the intake fan or, in single-fan setups, where the intake fan would have been.

Each heatsink was mounted three times. Each time it was mounted, the TIM was allowed a day to fully spread before the temperature tests were run. Shorter and longer times were tested. Shorter mount times resulted in hotter temps; longer times did not improve temps.

Each test run was 60 minutes in duration. The last 50 minutes of each run was measured, and the temperature logs were analyzed in Open Office spreadsheets. The ambient was measured every five seconds, resulting in 600-line spreadsheets. An Intel chip reports its temps in one degree increments, so for best accuracy these reports should be averaged in aggregate. Here, the core temps were measured once a second, resulting in 3000-line spreadsheets. The mean ambient temp was subtracted from the mean core temps, resulting in a net temp for each run. The net temps were then averaged. On the NH-D14, one of the three mounts resulted in consistently higher temperatures. Consequently, those results were not included in the calculations of temperatures so the results are an average to two runs each.

The sound pressure level was read by an SPL meter mounted on a tripod, one meter to the side of the heatsink. This basement sounds silent, but the SPL meter registers 30 to 30.5 dBA in this silence. The SPL of each heatsink could be reported as measured (31.5 to 41.5 dBA) or net (1.5 to 11.5 dBA). There are pros and cons for each method, but the as-measured method seemed to accord with a subjective impression comparing the noise with other fans heard at one meter. But the difference in noise was better captured by converting the SPL number to a net SPL by subtracting 30 from the SPL measurement. Thus, both numbers are included in the Result chart.

Recall that the NF-P14r redux-1500 PWM is the only type of NF-P14 with a round frame that is still sold. The retail NF-A15 has a nominal speed of 1200 RPM, so the NF-A14 FLX was an adequate stand-in. Both fans were tried with the stock NF-P14 PWM so you can have a basis for comparing price vs performance.

Using a pair of NF-A14s on both the NH-D14 and the NH-D15 allows for a direct comparison between the two heatsinks.

As for mounting your heatsink, you want the best mount for your heatsink, one that when you remount it, you will get no better cooling. The only way to do this is by trial and error: you use varying amounts of TIM until the cooling gets no better. Only then do you have a keeper.

The particular method used here to create successful mounts involved placing a 5mm bleb of TIM in the center of the CPU’s IHS (integrated heatsink, or the CPU’s top cover). The heatsink was then set down on the bleb and tightened. Immediately after mounting, it is clear the TIM has not yet spread to the far corners of the CPU because the system is more than a degree hotter. Give it overnight, though, and it is ready to go. The pictures below shows sample mounts from the NH-D14 and the NH-D15.

Note the bare patch in the corner. That indicates that the TIM did not swamp the interface between CPU and heatsink. The bare patch, BTW, showed up in different corners depending on which direction the bleb was a little off-center (no human endeavor is perfect). Otherwise, the TIM-prints indicated a uniform spread, with no significant irregularities.

The NH-D15, the NF-P14r and the two NF-A14s were provided by Noctua.

First of all, look at the chart below. It tells you that the NH-D15 cools better than the NH-D14, and is about as noisy. But the prize finding is on the second line of the chart. With the RC-7 Low Noise Adapters attached to the fans, the D15 cools about as well as the D14, and you can barely hear it. Remember that a silent room shows up as 30 dBA on an SPL meter. On the LNAs, the D15 is so quiet it is the next thing to silent.

Using either heatsink with a single fan did not reduce the noise, but greatly reduced the cooling.

The NF-A14/A15 fans are among the winners here. When the A15 stand-ins were used, temperatures went down and noise stayed low.

One issue with the OCCT testing is that it does not push the heatsinks as hard as they can be pushed. Preliminary testing with Linpack suggests the differences between heatsinks become more pronounced. But until the methodological issues are worked out, the results remain preliminary.

The NH-D15 did not appear from thin air. It had a predecessor to beat – a very good one. The requirement that the heatsink not be too tall or too wide set hard limits on the designers. Given those constraints, how could they make a better cooler?

Noctua improved on the Gold Standard of heatsinks, which until now seemed impossible. In working with this heatsink (remember, I had this mounted and dismounted several times and did a number of tests that were not included in the final analysis) I was impressed with three things. First, how much easier the NH-D15 is to work with overall. Second, how the D15 cools a little better than the NH-D14. Finally, the D15 is a whole lot quieter than the D14. Very quiet indeed.

The sweet spot for quiet lovers is to get an NH-D15 and attach Low Noise Adapters to both fans. Set up this way, the NH-D15 cools like a champ while literally purring like a kitten. Actually, since I have raised kittens, I can tell you that they purr much more loudly than the D15.

Noctua has evolved the design of their clips while maintaining backward compatibility – no mean feat. They have also improved their already world class mounting system. They have kept the top low enough to fit into standard cases. If it has a 120 mm fan on the back, the D15 will fit.

If you already have an NH-D14 and you want to improve your cooling, just buy a couple of NF-A15 fans. To make best use of any heatsink that uses a pair of 140 mm fans, make sure you buy low profile RAM. Buying RAM with coxcombs is wasting your money and compromising your cooling.

Finally, if you have ever worked with Noctua, you know they provide world-class customer service. Every once in a while someone on a forum needs something from Noctua, be it a replacement fan or a new set of clips. Their experience has left them uniformly pleased.

A note on weight: you cannot build a major league heatsink and keep it lightweight. Great cooling implies a certain weight of metal. If your motherboard is fastened with all of its screws, it will not deform under the weight of a large heatsink. And the NH-D15 is not the heaviest heatsink.

The Noctua NH-D15 is the new Gold Standard of air-cooling.

High quality costs – this is not a cheap heasink. It’s available at Amazon for $89.99 or Newegg for $84.99.

Click the stamp for an explanation of what this means.

Ed Hume (ehume)