Table of Contents
Whether they are for HTPC’s or simply for taking up less room on the desktop, products designed for slim form factor cases are in high demand. CPU heatsinks for cases like this typically blow air down on the motherboard, cooling not only the CPU, but the various components on the motherboard as well. Noctua has responded with five down-blowing CPU heatsinks. This review will focus on the middle model, the Noctua NH-L12. As usual for Noctua heatsinks, it is designed for flexible fan mounting. The fans are controlled by your motherboard through silent PWM circuits. So how well will it cool your CPU? Let’s find out.
The Noctua NH-L12 when fully decked out has a 120mm NF-F12 PWM fan on top and a 92 mm NF-B9 PWM fan on the bottom. The smaller fan allows room for four sticks of RAM, even RAM with tall heatsinks. The full assembly is 93 mm tall with the 25 mm-thick fan on top, 66 mm tall with the top bare. The heatsink itself is 118 mm long, measured in the direction the heatpipes run. It is 128 mm wide across the run of heatpipes, which allows the top 120 mm fan to snuggle into the center. Across the width of the heatsink, the fins are cut so they make the fin stack 17.4 to 18 mm thick where the 92 mm lower fan sets into the bottom. Its 23 mm thick on its outside edges. Roughly, then, the heatsink is about 120 mm square by 20 mm deep.
|Socket compatibility||Intel LGA2011 (Square ILM), LGA1366, LGA1156, LGA1155, LGA775 & AMD AM2, AM2+, AM3, AM3+, FM1, FM2 (backplate required)|
|Height (without fan)||66 mm|
|Width (without fan)||128 mm|
|Depth (without fan)||150 mm|
|Height (with fan)||93 mm|
|Width (with fan)||128 mm|
|Depth (with fan)||150 mm|
|Weight (without fan)||415 g|
|Weight (with fan)||680 g|
|Material||Copper (base and heat-pipes), aluminium (cooling fins), soldered joints & nickel plating|
|Fan compatibility||120x120x25mm & 92x92x25mm|
|Scope of Delivery|
|Model||Noctua NF-F12 PWM & Noctua NF-B9 PWM|
|Max. Rotational Speed (+/- 10%)||1500 / 1600 RPM|
|Max. Rotational Speed with L.N.A. (+/- 10%)||1200 / 1300 RPM|
|Min. Rotational Speed (PWM)||300 / 300 RPM|
|Max. Airflow||93,4 / 64,3 m³/h|
|Max. Airflow with L.N.A.||74,3 / 52,6 m³/h|
|Max. Acoustical Noise||22,4 / 17,6 dB(A)|
|Max. Acoustical Noise with L.N.A.||18,6 / 13,1 dB(A)|
|Input Power||0,6 / 0,96 W|
|Voltage Range||12 V|
|MTBF||> 150.000 h|
Noctua waited a long time to begin fitting its heatsinks with PWM fans. Their reason is that most PWM fans cause periodic small distortions that result in a clicking noise. Noctua’s engineers were able to devise circuitry to eliminate this annoying noise. The new circuit design allowed Noctua to produce “no clicking” across their entire speed ranges (see the review here). Another feature of Noctua PWM fans is an approximately linear fan response to PWM signals. A 25% PWM “duty“ produces about a 25% speed, 75% duty produces about a 75% speed, etc. Here we see the 92 mm NF-B9 PWM and the 120 mm NF-F12 PWM. Their power cords are 200 mm or 8 inches long. If you need a longer reach, Noctua provides one.
The NH-L12 box shows a Noctua-standard brown and white theme, so you can tell at a glance which company is selling this heatsink. Following a modern ethic, Noctua sells their heatsinks in all-cardboard packaging. You can recycle every bit of the package and packing materials, except for one little rubber piece that keeps the NH-L12 from being squeezed in shipping. The heatsink is held snugly in the package and sitting on top is the is the accessory box.
Removed from its packing, the NH-L12 reveals itself as a compact device. Unlike the NH-D14, the NH-L12’s fan clips do not require push-through pegs. They simply clip on the fans. The fans then clip onto the heatsink, where they can be removed and reinstalled quite easily.
Scope of Delivery
Arrayed on top of the three instruction sheets (AMD, Intel LGA 1366/115x/775, Intel LGA 2011) we can see the contents of the box. We start on the left with the Y-Split Cable. It is a 4-wire PWM cable, with one branch lacking an RPM-reporting line. This avoids confusing the motherboard with more than one RPM signal. Between the branches of the Y-cable is a tube of Noctua’s excellent NT-H1 thermal compound (TIM). Next is the NF-B9, a bag of AMD mounting hardware, the 30c m/one-foot PWM extension cable, an L-shaped Phillips #1 screwdriver (the short leg of the L is the handle), and two LNAs (Low Noise Adapters). Then we have the NF-F12 and the heatsink, a bag of Intel mounting hardware that includes both the SecuFirm2 Mounting Kit and the Mini-ITX Mounting-Kit. Up front there are four silicon rubber fan holders (a.k.a. “vibration isolators”) and four standard case fan screws. You only need four of either sort because at least one of the fans will be clipped to the heatsink. Finally we see the Noctua Metal Case-Badge.
The inclusion of Low Noise Adapters give this cooling system extra flexibility. If you do not wish your PWM fans to spin up to full speed, the LNAs will allow you to quiet these fans.
Mounting the NH-L12
Looking down on the NH-L12 heatsink, we can see holes in the fin stack that allow a clear view of two Phillips #1 screw heads. These are the tension screws. The second image shows the Noctua-provided screwdriver set into one of the tension screws.
By the way, this is the same mounting system used on other Noctua heatsinks. The NH-D14 that was compared with the NH-L12 mounted to the same hardware. It made swapping the two heatsinks a breeze.
The base plate is labeled for LGA 1366 LGA 1156 and LGA 775. It also fits the LGA 1155 and the upcoming LGA 1150. Two screws are placed correctly. The instructions warn you about incorrect placement. The second picture shows you a mounting backplate ready to put on the back of your motherboard. The pads are just the slightest bit sticky, to hold onto a vertical motherboard. This is a handy feature.
The left picture shows a SecuFirm2 assembly put together using the mounting plate on the bottom, the four mounting bolts, the four spacers, the two mounting bars, and the four thumbscrews. Your motherboard will sit on the four pads, while the spacers and the mounting bars will sit above the motherboard.
The right picture shows how you will mount the heatsink on the mounting bars. As Noctua says, you tighten the tension screws until they stop.
Before we get too deep, we should look at how SecuFirm2 interacts with a motherboard. In the first picture below is the mounting plate on the back of a motherboard. The bolt heads are captured by the lip of the backplate so they will not turn. Note the allowances made for the protruding caps of the socket’s screws. Make sure the backplate is properly positioned.
The second picture shows the top of the hardware on the other side of the motherboard. The lower two through-bolts are bare. The upper two are capped with Noctua thumbscrews. If you try hard you can put the mounting plates slightly askew. Make sure the through-bolt positions are symmetrical.
Now let us look at the base of the heatsink. As you can see, it does not have a mirror finish nor is it flat. The surface is subtly convex. Noctua has its reasons for this and they believe this is a good idea, as we shall see.
The left picture shows the heatsink tightened down. With the open end pointed toward the right, you can see how it overlaps the RAM. A fan on top of the heatsink at this point will directly cool the RAM.
|CPU||Intel i7 860 HT enabled, LLC enabled|
|Motherboard||GA-P55A-UD3P (sitting in open air or in modded case)|
|RAM||G.Skill Ripjaws DDR3-1600 @ 10x (stock) or 8x (overclock)|
|Graphics Card||PowerColor AX3450 Radeon HD 3450 (fanless, hence noiseless)|
|Solid State Drive||Kingston SSDNow V+100 64GB|
|Power Supply||SeaSonic X750 (fan mostly doesn’t run)|
|Heatsinks||Noctua NH-L12 and NH-D14, with Noctua NT-H1 TIM|
|Stress Software||OCCT 3.10|
|Tenma 72-942 Sound Pressure Level Meter|
|Digital TEMPer USB Thermometer|
In order to get a smooth heat curve, I used OCCT’s native CPU stress testing software. Those of you who have used Linpack based stress testing utilities know that Linpack seems to pause to take a breath now and then, and the CPU temperature falls. Not so with OCCT’s stress tester. With small data sets, it produces an even stress on the CPU that never falters until the run is complete.
When the NH-L12 is installed, I noted that it took a long time to reach thermal equilibrium. In order to assure the CPU and heatsink had reached a steady state, each run was allowed to go 40 minutes before the temperatures were recorded. The ambient temps and the core temps were averaged for the next 30 minutes to provide the results of each test run. In total, each test run took 70 minutes. Each test series would entail at least six test runs.
The NH-L12 and the NH-D14 were tested at the following overclocks:
Based on my earlier experiences with the NH-D14, I hoped to be able to test the overclocked CPU with the NH-L12 at 0.2 GHz intervals. That proved impossible, so the 3.7 GHz and 3.76 GHz were approaches to an overclock of 3.8 GHz. The high-end NH-D14 heatsink, used for comparison in this review, of course had no problem with overclocks of 3.8 and 4.0 GHz as it is widely considered one of the best air coolers on the market. For those not familiar with the product, it is substantially larger than the NH-L12 and quite a bit more expensive.
This NH-L12 spent a long time in the open air testbed. I record the performance of a heatsink as Temperature Over Ambient (TOA), the net temperature that remains when you subtract the average ambient temperature from the average temperature reported by the CPU. But unlike past experience, I simply could not get repeatable results. After a while I noticed something during the repeated test runs: most of the variability was not in the CPU temperature readings. The CPU temps were actually pretty steady for similar overclocks. But the ambient temperatures were all over the map, so the TOAs were not repeatable. This required a change in method.
I decided to measure the ambient temperature under the test stand, away from the heatsink’s intake. Surprisingly, that worked. I was now able to repeat the results of one test run with the results of another and get nearly identical TOAs. Success! But weeks were wasted. At six 70-minute runs per series, you can see the problem. Now it was solved though.
Initial exploration of the NH-L12’s cooling powers were done on an open air testbed. Then it was placed in a Zalman Z11 Plus case that was modded to remove its grills and approximate an open testbed, but one that would allow the heatsink to be tested on a vertical motherboard. Results in the case were re-checked to assure consistency. Note that the heatpipes run horizontally, front to back.
The NH-D14 and the NH-L12 were tested horizontally and vertically with their heatpipes running front to back. And then the NH-L12 was rotated 90 degrees and tested again, horizontally and vertically. The first picture shows the NH-D14. The second shows the NH-L12 with heatpipes running vertically. The bend in the heatpipes is below the heatsink. Noctua recommends this position. They advise against putting the bend of the heatpipes above the fin stack.
Before we get to the results, a few matters. With this test setup, if the CPU’s report temperature reaches 87 °C, the test run ends. With an ambient generally running 20-22 °C, we should not expect to see TOAs over 66-67 °C.
The NF-F12 PWM push fan usually ran around 1450 RPM. The NF-B9 PWM pull fan ran about 1700 RPM when it was with the NF-F12, and around 1780 RPM when it ran alone.
The sound pressure level (SPL) of the NH-L12 was measured from the side for comparison with other heatsinks, especially the NH-D14. However, since this is a blow-down cooler, the face of the intake fan and the flat part of the heatsink will face the side panel of your case, where you will normally have a grill (and perhaps a filter). So the cooler’s SPL was also measured at the face. Sound measurements were taken at 10 cm, with a 20 dB correction made to approximate the noise at one meter.
The NH-L12 teetered between quiet and moderate in its noise. What noise there is, is not harsh.
Cooling Results, Open Testbed
Let us look at the results of the three fan configurations possible with the NH-L12: Push-Pull, Push Only and Pull Only. Looking at the cooler with two fans, there seems to be little room under the bottom fan for air to get out. It seems possible that having a bottom fan will worsen the heatsink’s performance. This is something we must test.
The chart below shows test results on the open-air testbed specified above. Each test series stopped where it did because in each one, the last test would abort when the CPU temperature hit 87 °C. So anything faster threatened to overheat the CPU.
The pull-only NF-B9 PWM was able to cool the CPU with the NH-L12 at the stock 2.8 GHz and at 3.2 GHz. Beyond that it was too hot. The push-only NF-F12 PWM was able to cool the CPU nearly as well as both fans together. Not bad. Together the two fans enabled the NH-L12 to cool a hot-running first generation i7 CPU with a moderate overclock.
To get ready for testing in the case, let us compare the results on the testbed and in the case. The differences range from 0.5 °C to 3.8 °C, all in favor of the open test bench. With a case as open as this one, finding this much difference is instructive.
In testing the NH-L12 with its heatpipes running front to back with the motherboard vertical, all three fan setups were unable to cope with an overclock of 3.4 GHz. They ran too hot. Recalling that the fans were able to cope with moderate overclocking when the heatsink was horizontal, I added a fourth line, the push-pull configuration from the horizontal setup. When you consider that closing up a case will worsen the temperatures, it appears that the NH-L12 will only cool a stock chip adequately. No overclocking … at least with this configuration.
How good is the NH-L12? Let us compare it to the NH-D14, Noctua’s flagship heatsink. The top line in the chart below shows the NH-L12 with heatpipes running front to back in the modded case with the motherboard horizontal. The middle line shows the NH-D14 with heatpipes running front to back in the same case with the motherboard horizontal. The lowest line shows that NH-D14 with the case on its feet, with the motherboard vertical. The CPU speeds are 2.8, 3.2, 3.4, 3.6, 3.7, 3.76 and 4.0 GHz. For the NH-L12 with heatpipes running front-to-back, the horizontal position is the best case scenario. It is comprehensively beaten by the NH-D14, 10 °C at 3.76 GHz. At 3.8 GHz the CPU ran too hot to be cooled by the NH-L12, so the test was aborted. The NH-D14, of course, cooled better at 4 GHz than the NH-L12 at 3.6 GHz. Of course, we expect that kind of performance from the D14.
In the interests of being thorough, I reinstalled the NH-L12 with heatpipes running top to bottom. In the interests of brevity, only the push-pull results are shown. Those results are strikingly different from the results where the heatpipes run front to back. The top line shows the front-to-back orientation, vertical in the case. The NH-L12 could cool only to 3.2 GHz. After that the CPU got too hot. When oriented with heatpipes top to bottom and the motherboard horizontal in the case, the NH-L12 was able to handle an overclock to 3.7 GHz before overheating. Surprisingly, when the case was set up vertically, the NH-L12 was able to cool to 3.76 GHz before overheating. Starting at 3.6 GHz, the NH-L12 actually cooled better when it was vertical, just like the NH-D14.
Although there have been discussions about the effects of orientation on heatpipes, one would not expect a heatsink with heatpipes running vertically to out-perform a heatsink with heatpipes running horizontally. But that is exactly what we saw. Another interesting bit: when the NH-L12 was oriented with heatpipes front-to-back, it cooled better when the motherboard was horizontal. But when the NH-L12 was oriented with heatpipes top-to-bottom, it acted like the D14 and cooled better when the motherboard was vertical. Clearly, this had nothing to do with heatpipe orientation and everything to do with the interaction of heatsink and CPU. The orientation of the heatpipes is just a marker.
Looking for an Explanation
We will look for an explanation of these results by first looking at the relationship between the heatsink and the CPU. The heatsink sits on the CPU with a layer of TIM (thermal interface material) in between. This layer ideally should be microscopically thin, not fat as the drawing below shows. But the drawing below shows the relationship.
A “CPU” contains other pieces, though. The cover with which the heatsink interacts is really the IHS (integrated heat spreader). Between that and the actual CPU chip is another layer of TIM. In an Ivy Bridge chip like the i7 3770K, the TIM is an elastic substance. In an i7 860 like mine, this TIM is a low-melting-point solder. With the inner TIM being solder in our setup, there is nothing that changes with the heatsink orientation. So we should look at the interface between the CPU and the heatsink for an explanation of the variation.
First of all, we will set up our pictures to show the orientation of the NH-L12 and the motherboard. In the left image the PURPLE arrow shows the direction of the heatpipes. In the second, the RED arrow shows the direction from the RAM to the CPU, along with the purple arrow showing the heatpipes running front to back.
Let us read tea leaves now and look at the TIM patterns on our CPU and the contact face of the NH-L12. In the first series of tests where the heatpipes ran front to back, the results were not as good as the second series of test runs, especially when the motherboard was vertical. The pattern we see is that the center has little TIM. We should expect that the best heat conduction would occur there, where the NH-L12 and the CPU made their best contact.
In the second set of test runs, when the heatpipes were oriented top to bottom, we see a different pattern: there is an elongated patch of good contact. That patch is a lot bigger than the first patch.
Let us strip off the TIM and look at the CPU’s IHS, or integrated heat spreader, in its clean state. This is the copper cover of the chip that takes the heat from the CPU, spreads it out and transmits it to the contact face of your heatsink. This i7 860 has seen many a mount, and shows it. And the pattern of wear is interesting. It looks to show a chip under the IHS. So, let us mark it in the second picture.
Now let us look at the contact surface of the NH-L12. From the two pictures we can see it is convex, with the curve more pronounced running across the heatpipes. So our convexity is elongated and aligned with the heatpipes.
Let’s mark this, too.
Now our results make sense. In the test runs where the heatpipes ran front to back, the oval and the rectangle crossed, resulting in a smaller patch of close contact.
In the test runs where the heatpipes ran top to bottom, the oval and the rectangle were aligned with each other, producing a larger patch of good contact.
Well, this heatsink was certainly put through its paces. I have spared you the details of the remounts to confirm the earlier findings. The NH-L12 held up well.
Noctua’s NH-L12 is a finely crafted heatsink that is designed to fit in small places. Even at its lowest profile with a less-than-optimal contact, this 66 mm CPU cooler will cool an i7 860 at stock speed. If it will cool that chip, it will certainly cool your non-overclocked CPU.
We learned that this cooler has limits on its ability to cool an overclocked chip. Depending on your case, the fan setup you choose and the goodness of fit between the heatsink and the CPU, the NH-L12 may or may not handle a mild to moderate overclock. However, if you are looking to overclock your rig you should look to a different heatsink. The more expensive ($20 more) and much larger NH-D14, for example, was designed for that task.
Finally, you should test this heatsink in two orientations 90° apart to find the setup that suits your system the best. It appears that some orientations are better than others. And yet it is hard not to get a good mount. Along with the SecuFirm2 mounting system, the convex contact surface is designed to assure excellent contact between the NH-L12 and the CPU. It does that. While some orientations are more excellent than others, if you use the right amount of TIM (enough but not too much) you will get an effective mount.
There are other fine aspects of the NH-L12 that you find with all the Noctua heatsinks: the SecuFirm2 mounting system, for example, is easy to use. The package includes all the accessories you need, from a Y-cable to speed-reducing wires to a very clear set of instructions. The NT-H1 TIM is easy to work with, and has performed well in testing by a number of reviewers. The fans that come with the NH-L12 are both PWM and include the now-famous NF-F12.
It is all packed in recycling-friendly packaging, but who would want to throw it away? The cardboard boxes go back together easily so you can store the box on your shelf to await the time you may move, or wish to store the NH-L12 for your next project.
- Exquisite craftsmanship throughout.
- Clear instructions provided.
- Even with a less than optimal contact, this heatsink will cool a non-overclocked CPU.
- Convex baseplate contact surface assures good contact with IHS.
- Easy and secure mounting through the SecuFirm2 mounting system, which holds tight.
- Many options for setup – you get to set it up the way you like it.
- Highly rated thermal interface material is included.
- World class customer service.
- Little headroom for overclocking (though to be fair, heatsinks of this design are not meant for overclocking).
- Second fan does not always improve cooling.
The Final Word
This is not a budget heatsink ($69.99 on Newegg). Because of the care and quality that goes into the NH-L12, it costs more than other heatsinks that have limited cooling capacity. You get what you pay for.
– Ed Hume (ehume)