by Sergey Lepilov
04/17/2006 | 04:09 PM
Just what do you know about Austria anyway? Mozart, the soccer club Sturm, the ex-F1 racing track A1 Ring, superb skiing resorts, and Gцsser beer. Well, that’s about all I could recall at the moment. Of course, there’s a lot more fascinating things about Austria, but you will hardly produce a name that has anything to do with computer coolers even if you think hard. We’ve all got used to the fact that manufacturers of cooling systems mostly reside in Taiwan, Malaysia, Japan and, a few of them, in the USA. That’s why we were somewhat surprised to get two heatsinks on heat pipes from an Austria-headquartered Noctua for our tests.
Before I got down to test the received coolers, I dropped in on the Noctua website and learned that the company had been founded as a joint venture of the Austrian Rascom Computerdistribution Ges.m.b.H and the Taiwanese Kolink International Corporation (so, it’s not absolutely without Asians anyway).
Noctua was formed on the basis of the Austrian Institute of Heat Transmission and Fan Technology, and this looks like a guarantee of a very serious, scientific approach to the subject. By the way, a Noctua is a little owl, the symbol of Athena, the goddess of wisdom, sciences and strategy. Noctua Company sees its mission in putting intellect above brute force and embodying this concept in its products.
Getting closer to the actual products, the official website currently names two models of cooling systems for the CPU: Noctua NH-U9 and Noctua NH-U12. I will examine and test them both in this review.
The Noctua NH-U9 and NH-U12 heatsinks are shipped in packages of the same design:
The back side of each box gives you a brief list of tech specs, a list of accessories included, and a couple of photographs of different parts of the heatsinks.
There is a special compartment in each box the size of a 120mm fan in the NH-U12 model’s box and of an 80mm or 92mm fan in the NH-U9’s. So, although the heatsinks are currently shipped without fans, the manufacturer has provided for that as a future possibility.
The heatsinks are firmly fixed inside the boxes to minimize the risk of damage during transportation:
Lower still, at the very bottom of the box, there are the accessories and a small manual on installing the heatsink on the supported CPU sockets in 4 languages (German, English, French and Spanish).
And here are the accessories supplied with the coolers:
Besides the manual book, the following components are to be found in the small plastic bag:
The Noctua coolers cannot be installed on Socket 462 (Socket A).
The coolers represent the already classical “tower” design: a copper foundation with heat pipes going through, and aluminum plates on the pipes.
What’s externally different between these two coolers is their size. The number of aluminum plates and heat pipes, and the base area are absolutely the same, so I am going to describe the senior, NH-U12 model in detail, mentioning special features of the junior NH-U9 where necessary.
So, we’ve got 38 aluminum plates here and the dissipation area of the NH-U12 is over 5000sq.cm (as opposed to the NH-U9’s 3800sq.cm).
The aluminum plates are the same thickness of 0.4-0.5mm in both the models. The senior model is 29mm taller, so the plates are placed at a different distance: about 2mm in the NH-U9 and about 3mm in the NH-U12. It may mean that a faster fan may be necessary for the Noctua NH-09 whereas the NH-U12 can get along well enough with a low-speed fan. By the way, Noctua puts an emphasis on that issue in its press release about the reviewed coolers, saying that both the heatsink models are highly efficient even with a low-speed fan. We’ll of course check this claim in our tests shortly.
The ends of the heat pipes are neatly soldered up at the top. A few slits and holes of unclear purpose can be seen nearby. I suppose they were used when the heatsink was being manufactured or assembled.
You can see in this snapshot that the edges of the aluminum plates are profiled, just as I’ve seen in coolers from other manufacturers. This profiling is meant to minimize resistance to the air flow and to increase the overall cooling efficiency. The copper heat pipes are offset a little relative to one another – I think it is not because the heatsink design demands so, but because the manufacturer wanted to distribute the thermal load uniformly in the heatsink’s plates.
Note also that the middle part of the heatsink’s side plates is curved in to narrow the air flow and drive it along the plates rather than sideways. The engineers have really paid attention to every small detail and reduced airflow loss to the minimum in this design.
There is a sticker on the cooler’s sole to protect it from scratches:
The cooler’s bottom is finished just perfectly:
There is no trace of soldering or thermal glue where the heat pipes have contact with the cooler’s copper sole:
But the bracket above the pipes is indeed soldered down to the base:
The copper base of the Noctua coolers is 7.5mm thick; its area is about 17.5sq.mm. Four threaded holes are located around the perimeter of the base. They are used to install a fastener and mount the cooler on the appropriate platform. The heatsinks are rather light by today’s standards, weighing 570g (Noctua NH-U9) and 700g (Noctua NH-U12).
Remove the protective film from the cooler’s sole before you install it! Don’t snub this remark as I myself have lost count of how many times I had to correct this problem after inexperienced users turned to tech support crying they had been given a defective cooler. :)
It was easy and simple to install the heatsinks. You should fasten two U-shaped plates to the heatsink’s sole to mount it on a Socket 939/754/940 mainboard.
Then put a thin layer of thermal paste on the processor. Next, install the heatsink on the CPU’s heat-spreader and fasten it to the appropriate back-plate with two black spring-loaded screws.
Note that there’s nothing at all at the heatsink’s left side, but there are memory modules under its right edge. I actually mounted the heatsink without problems, but you may find it difficult to plug memory modules in and taken them out with the heatsink installed.
To complete the installation of the cooler on Socket 939/754/940, hitch a 120mm fan to it using the wire braces.
The orientation of a Noctua heatsink on a Socket 939/754/940 platform depends on the position of the mounting holes as well as of the CPU socket itself. If they are placed along the mainboard, the heatsink cannot be oriented in the best way possible, i.e. so that the fan was blowing towards the back panel of the system case.
The Socket T (LGA775) platform is free from this drawback. Before mounting a Noctua cooler on a Socket T (LGA775) mainboard, glue the appropriate back-plate to the mainboard’s reverse side (do not forget to remove the protective film from the back-plate before doing that).
Then you should use the included spacers to fasten two rails to the back-plate through the mainboard’s mounting holes.
The holes are placed in a symmetrical order around the CPU socket on any Socket T (LGA775) mainboard, so these rails can be placed in any position to orient your Noctua in the best possible way inside the system case. A couple of shorter rails are used for Socket 478. Note that positioning the rails as shown in the snapshot will not prevent you from putting the CPU into the socket later. It is harder to insert the CPU or take it out if the rails are placed in parallel to the socket lever.
Having fastened the appropriate brackets to the heatsink’s sole, you should then secure the heatsink on the CPU with two light spring-loaded screws.
And finally you have to put a fan down on the heatsink:
As you see, there’s no obstacle to mounting a Noctua NH-U9 on an ABIT AA8-DuraMAX mainboard, but if you tried to install the larger Noctua NH-U12 (which is 29 millimeters wider), the chipset cooler would get in your way. You can avoid this difficulty by simply turning the chipset’s cooler round so that its airflow was directed towards the CPU socket. I can’t say how compatible the Noctua NH-12U is with other mainboards, but the distance between the cooler’s sole and the bottommost plate of the heatsink is 32 millimeters. I guess this information may be useful to you.
To test the Noctua NH-U9 I used a 92mm Coolink fan (model 922) with a speed controller and a speed range of 1100-1850rpm (this is what the specs tell you, but the maximum speed of this fan proved to be about 2000rpm in practice). The performance of the fan varies within a range of 42 to 64CFM.
The Noctua NH-U12 was tested with a 120mm fan from Thermaltake (Thunderblade A1926) with a speed controller from the Coolink fan (a speed range of 1100-2000rpm and a performance of 42-78CFM).
Here are the tech specs of the coolers about to be tested:
Thermaltake Big Typhoon
38 aluminum plates
2x70 aluminum plates
Dissipation area, sq.cm
Cooler (fan) dimensions:
70 x 95 x 128
70 x 124 x 155
122 х 122 х 103
Nominal voltage, V
12 (7 - starting)
Power consumption, W
Fan rotation speed, RPM
~ 1300 (±10%)
1, slide bearing
Supported CPU sockets
Socket 462 (A)
Maximum supported CPU
All Intel Pentium 4 and AMD Athlon 64 processor models
I tested the coolers on the following testbed:
The tests were performed in Windows XP Professional Edition Service Pack 2. I disabled the fan speed management options in the mainboard’s BIOS.
To make the test tougher for the coolers, I overclocked the AMD Athlon 64 3200+ processor from its default 2000MHz to 2700MHz frequency at 1.6V voltage.
The temperatures and fan speeds were monitored with the help of SpeedFan version 4.28:
I used the S&M version 1.8.0 (alpha) utility to heat the CPU up for 15 minutes with the FPU test at 100% load.
S&M puts an unrealistically high load on the CPU, so I also tested the coolers with SuperPI by calculating pi to 32 million decimal places; this task takes a little less than 29 minutes on the described testbed. I guess this is enough to evaluate a cooler’s efficiency. The temperature was read from the sensor integrated into the CPU. The room temperature was 22-23°C during the tests and is taken as the initial point in the diagrams.
I will compare the performance of the Noctua coolers with that of one of the best air coolers available, Thermaltake Big Typhoon (for details you can check our article called Four CPU Coolers from Thermaltake Tested). I also offer you the results of a standard boxed cooler from the AMD Athlon 64 3200+ processor for comparison:
I first checked the efficiency of the heatsinks’ native thermal paste. The syringe hasn’t any identifying marks or labels:
I contacted Noctua and received an immediate and “exhaustive” response that “Noctua heatsinks come with a highly efficient silicone-based thermal interface”. Yet the company representative also stressed the fact that the Noctua coolers perform the best when used with Arctic Silver 5. Anyway, the native thermal paste included with the heatsinks is quite enough and I used it in my tests.
The Noctua NH-U12 and the Thermaltake Big Typhoon (also with the above-mentioned Thermaltake Thunderblade A1926 fan) were tested at two fan speeds: a quiet 1280rpm and a maximum 2000rpm. The Noctua NH-U09 was tested with a 92mm fan at a very quiet (almost silent) speed of 1600rpm and at the maximum speed of 2000rpm. I made pauses of about 30-40 minutes between the test cycles for the temperature to stabilize.
Let’s first look at the results I’ve got in a closed, but well-ventilated system case:
So it is clear there’s now one cooler more which is better than Thermaltake’s Big Typhoon. To be exact, the Noctua NH-U12 keeps the CPU cooler by 2°C than the Big Typhoon at both the min and max speed of the fan under the S&M load. The CPU temperature goes down by 4°C with both these coolers when the fan speed is increased to the maximum. However, this is only true for S&M (or Sado-Maso as it is sometimes referred to by overclockersJ) whereas the temperature difference under the Super PI load is a mere 1°C.
You may be wondering why buy a super-cooler at all if the ordinary boxed cooler from AMD can cope with a processor overclocked to 2.7GHz at increased voltage and running S&M? Well, just look how fast its fan is! The fan is too noisy at 4700rpm, you will hardly want to be near this computer for long. Even the noisiest of the fans I’ve tested today (Thermaltake’s Thunderblade A1926) is not as hard on your ears at its 2000rpm as the boxed cooler’s fan is.
As for the junior NH-U9 model, it is not far behind the Noctua NH-U12 in the quiet operating modes, but at the max fan speed there is a difference of 5°C between them. The Noctua NH-U9 seems to ask for a faster fan since the 92mm is less noisy at its maximum speed of 2000rpm than a 120mm fan at the same speed. In other words, the junior model loses in terms of performance but wins in terms of noiselessness under such test conditions. Still I would rather judge a cooler by its efficiency in quiet modes which are more relevant for practical use.
Now I’m going to eliminate the influence of the system case on the test results and see what happens on an open testbed. In this test I put two 120mm Sharkoon Luminous Blue LED fans (1000rpm, 21dBA) on both sides of the Noctua NH-U12 heatsink for driving air in and out.
Note that neither of the Noctua coolers needs a high-speed fan to pass the S&M test on an open testbed. It’s only in SuperPI that the higher fan speed improves the temperature; there is no such difference under S&M. So I can’t but recall the claim of Noctua that its coolers don’t need a high-speed fan – this is indeed so! The Thermaltake Big Typhoon, on the contrary, reacted eagerly to the higher fan speed. Its efficiency got 2°C higher and it became one of the leaders in the open testbed tests.
My attempt to use two 120mm fans on the Noctua NH-U12 (one for driving air in and one for driving it out) proved to be a failure. The fans seem to have been hindering one another rather than helping.
Acknowledging the highest efficiency of CPU cooling, I still can’t pass by one drawback. The temperature reported by the PWM sensor was always higher in my tests of the Noctua coolers than with the Thermaltake Big Typhoon; it is even higher than the CPU Sensor showings with the NH-U9. This is expectable since the Big Typhoon directs its air stream towards the mainboard’s PCB while the Noctua coolers, in parallel to it.
Our today’s tests have produced two more super-efficient CPU coolers ready to cope with the huge amount of heat produced by overclocked processors. Both the Noctua coolers boast an impressively high CPU cooling efficiency with little dependence on the fan speed which is an indication of an ingenious and well-implemented design of the heatsinks. High efficiency, convenient fasteners, simple installation, good thermal paste – what else could an overclocker wish for? You may answer – moderate pricing, but it’s all right with this factor, too. The official price of the Noctua NH-U12 and Noctua NH-U9 is 47.9 and 37.9 euros, respectively. And here are their advantages and drawbacks summed up for you once again: