Noctua NF-A12x25 PWM: The most fine-tuned fan of its kind

Noctua NF-A12x25 PWM in detail

All the glory and respect of the Noctua NF-A12x25 “sterrox” fan is well deserved. We waited a long time to release its tests, mainly because it was necessary to analyze more seemingly similar designs first. This is necessary to understand why Noctua’s most popular fan among standard thickness models has no competition. We can responsibly say that even though we have only tested a fraction of all the fans.

We know which fan’s results you’ve missed (and perhaps missed the most) in tests so far. Anyway, if we had released the Noctua NF-A12x25 tests earlier, the ranking wouldn’t be as clear as it needs to be. To get a good understanding of why this fan is really the pinnacle in terms of aerodynamics, it’s important to understand the shortcomings of other models. That’s mainly why in the other four tests we focused on fans whose blades were similar in shape, but still with often significantly different results. We’ll go through all of that during this analysis and explain where Noctua had to go the extra mile to make the NF-A12x25 shoot to the top ranks of aerodynamic efficiency.

Because of its undeniable visual similarity, the Noctua NF-A12x25 fan immediately earned the “Gentle Typhoon copy” label right after its introduction. In reality, however, Noctua only took inspiration from the shape (much like Scythe, which once saw it elsewhere and applied it to the Gentle Typhoon), which would be rather silly to ignore. It would be similarly foolish to someone trying to forcefully come up with a “new” wheel that had a geometry other than circular. So yes, Noctua has acknowledged that the basic shape of the Gentle Typhoon is more efficient than designs with less curved leading edges, and there’s no point in denying that. At the same time, it’s nothing that can’t be improved upon quite significantly.

Akasa has also jumped on the “à GT” fan bandwagon with the Alucia SC12. However, as you already know, that is not one of the most efficient fans. This is because a very flexible material is used to make the rotor, which significantly reduces static pressure. The airflow in a non-restrictive environment is excellent with respect to noise level, but its losses due to obstacles are more significant than in similar fans with more rigid blades (and the efficiency, the ratio of airflow to noise level produced, is also significantly worse). And why is a design with more aggressively curved blades actually more efficient than designs with less curvature on the leading edges? To direct the air currents. The greater the curvature of the blades, the smaller the flow angle. From a certain point (extreme curvature) onwards, this can reduce the efficiency of the flow just as much as it reduces it when the active flow angle is too large. Aside from the unwanted turbulence created at the edges of the frame, for the same airflow, higher friction of the currents against the stator frame increases noise (and reduces efficiency).

With a rotor design like Noctua’s, there is not such a high concentration of air currents in these locations, friction is lower (and with that comes lower noise), and maximizing axial airflow greatly increases efficiency. Therefore, we can confidently rule out that a structurally identical fan (i.e. also made of identical materials) would achieve the same air with, say, a traditional rotor shape (with small curved leading edges) at a lower noise level. This would only be possible, for a number of reasons, if a rotor with a more aerodynamically efficient shape was made of a less rigid material (and thus more subject to unwanted deformation during rotation). The basis of the NF-A12x25’s success is therefore “sterrox” with extremely high compressive as well as tensile strength (unlike the Corsair AF120), and thermal expansivity is also lower than cheaper materials.

But what actually is this “sterrox”? LCP (Liquid Crystal Polymers) with something secret mixed in, which in sum refers to Noctua’s patented sterrox. But even “pure” LCP already achieves very high strength. The MSI MEG Silent Gale P12 is also made of it, and usually in the top 3 fans tested so far, it holds its own even through obstacles. It lags behinds the NF-A12x25 mainly due to less emphasis on aerodynamic details, which Noctua has prided itself on since its inception. Before we get to them (and their contribution), looking at the rotor here offers one comparison to the BeQuiet! Silent Wings (Pro) 4.

Note that the NF-A12x25 has significantly larger gaps between the blades. Naturally, these reduce static pressure, but that doesn’t mean that a fan with smaller gaps and more total blade area (which the Silent Wings 4 also has) has to achieve higher static pressure. Again, among other design details that affect the overall static pressure, the material used matters. BeQuiet! uses PBT, which although reinforced with glass fibers, is still more flexible on the Silent Wings (Pro) 4 with its relatively small thickness (especially in the weakest spots between the grooves) than LCP. That’s what degrades the performance of the new BeQuiet! fans the most and it also causes undulations, the occurrence of more unpleasant frequency spikes in the sound and also mechanical vibrations, which transfer in higher intensity from the blades to the frame (and then to the case and so on).


Noctua eliminates these undesirable phenomena on the NF-A12x25 by using extremely strong sterrox, which, moreover, has not been skimped on. The blades are also really thick. Thanks to this, Noctua was able to reduce the gap between the tips of the blades and the frame to some 0.5 mm. In the vast majority of other fans, this gap is two to four times larger. This is naturally due to, among other things (around precision manufacturing with low tolerances), blade instability on the run. And while this is a remarkable static pressure-increasing detail, it should be remembered that of the total cross-sectional area, this air-leakage gap is a relatively small fraction. Therefore, this feature may need to be given more weight than it really has.

We expect that as part of the next generation of fan improvements, Noctua will focus more on reducing the gaps between the blades (so that the leading edges of the blades are closer to the trailing edges of the previous blades). Finally, the space between the blade tips and the frame can’t be improved much aerodynamically anymore. The increasingly popular framing (for example, as with the Cooler Master MasterFan SF120M) is suitable at most as a compromise for cheap fans that have to skimp on lower-strength materials.

It suppresses air leaks around the rotor much more elegantly. In addition to what we already mentioned (the rotor more pressed against the frame), there’s also the unconventional finish on the inner walls of the frame that Noctua introduced with the NF-F12 fan. The idea is that the surface of the frame is “roughened” by recessed triangle-shaped formations (Inner Surface Microstructures), these form a microturbulent layer which, in addition to the overall higher resistance (and thus the desired braking of the returning air by the back pressure), suppresses the break-off of air currents from the blade tips. This does not happen on smooth walls and therefore the static pressure and the volume of air flowing over the obstacle (whether it is a dust filter, a grille or a radiator) decreases. Compared to Noctua’s other fans, the NF-A12x25 ISMs have a smaller depth, which is directly related to the thinner gap between the blade tips and the frame. For the desired aerodynamic effect (eliminating airflow separation from the blade tips), while keeping noise as low as possible, less aggressive ISM dents are suitable, says Lars Strömbäck, the architect of this fan.

But Noctua has even more aerodynamic details to improve efficiency on the NF-A12x25 (PWM). One of them is the channels on the entry side of the exhaust sides of the blades, which are mainly to suppress unwanted microturbulence and thus accelerate the airflow to increase the “suction” capacity. They also partially prevent the air currents from sliding down the long side of the blades, up to their tips (this also reduces the efficiency of fans). The third important and within Noctua fans (but also in general) an exclusive feature are some kind of steps (Stepped Inlet Design) on the intake side of the frame. These are supposed to break up the sound of the circulating air with a view to reducing frequencies that should be more pronounced on normal/smooth frames, more unpleasant for the user.

And one more thing improves the aerodynamic properties of the NF-A12x25. This time, however, it is no longer a solid part of the design, but a silicone gasket that can replace the pre-installed anti-vibration pads in the corners. These are also included within this gasket, but unlike them, this gasket also fills the gaps that occur between the separate anti-vibration corners. This space naturally allows some air to escape and in turn reduces the static pressure, albeit it is not a dramatic difference. While without the gasket (only with anti-vibration pads in the corners), in a non-restrictive environment, we measured a maximum static pressure 2 % lower (2.29 mm H2O) than is specified (2.34 mm H2O), it is 2 % higher with the gasket (2.39 mm H2O). Of course, these ratios may also vary in practice depending on the mounting pressure.

To maintain the highest possible accuracy and philosophy of testing fans with pre-installed accessories (and not the best configuration for each of the measurement aspects, so as not to confuse matters), all measurements were taken with standard corners. Sooner or later, we’ll go through more situations with and without the gasket and I’ll also observe the effect of downforce with varying degrees of compression (and narrowing of air gaps) of the silicone material. But there will be a separate test for that. Now we expand on this topic mainly to give you an idea that the effect of this seal on static pressure and airflow is quite small, much smaller than the various corners on the BeQuiet! Silent Wings (Pro) 4 fans. And also because the coefficient of parameter truthfullness can have be of slightly positive or negative nature just depending on whether this gasket is used or you stay with the basic setup (with pre-installed pads).

Noctua talks about the sealing gasket in the context of liquid cooler radiators, but it also makes sense when mounting fans in system positions, although it will have very little effect on cooling performance. The biggest difference in performance with and without the gasket is at maximum power (about 3.8 %), at lower speeds it is always less, until the difference gradually approaches zero. We will certainly return to this topic in the form of more detailed measurements, probably as part of the enrichment of the NF-A12x25 chromax. black tests, whose results should match the brown-and-beige variants better, as is the case between the white and black Corsair AF120 fan variants.

We still haven’t quite covered the basics here yet though – the Noctua NF-A12x25 is a 120mm format fan with a standard profile thickness of 25 mm. The variant being measured in this test is “PWM“. In addition to it, Noctua also offers the “FLX” variant (speed also up to 2000 rpm, but only DC control), two slower 1200 rpm variants: “ULN” (with DC control) and”LS-PWM” and two with a lower nominal voltage of 5 V (again, one three-pin for linear control, and the other four-pin, with control by changing the width of the voltage pulses) capping officially at 1900 rpm.

Also worth mentioning are the above-standard plentiful accessories. Not many other companies give so many different things like Noctua for fans even in this price range. Specifically, it is a Y splitter (for connecting two fans to one connector), a 30 cm extension cable (with it you get to the final length of half a meter) a NA-RC14 deceleration adapter, rubber spikes for mounting in a case (as an alternative to self-tapping screws, which are also included in the package), and the aforementioned gasket.

* When reading performance values, a certain amount of tolerance must always be taken into account. For maximum speeds, ±10 % is usually quoted, minimum speeds can vary considerably more from piece to piece, sometimes manufacturers will overlap by as much as ±50 %. This must then also be adequately taken into account for air flow, static pressure and noise levels. If only one value is given in a table entry, this means that it always refers to the situation at maximum speed, which is achieved at 12 V or 100 % PWM intensity. The manufacturer does not disclose the lower limit of the performance specifications in its materials in that case. The price in the last column is always approximate.

Flattr this!

ARGB Asus fan (TUF TR120) now with a thicker profile

It’s evident from the design of the new TUF Gaming TR120 fan that Asus has diagnosed the weaknesses of its predecessor (and, from the current perspective, weaker model) well and worked on them with a view to achieving better results. These key changes are also easy to point out, and it’s likely that across all possible scenarios this will be one of the most aerodynamically efficient LED fans in the 120mm format. Read more “ARGB Asus fan (TUF TR120) now with a thicker profile” »


Arctic P14 PWM PST CO or ball vs. fluid bearings

Longer life in exchange for more noise? These are also some of the agenda items we’ll cover in our comparison of the Arctic P14 CO fan with the fluid bearing variant (P14). These are actually the main points. In any case, the ball bearings in the more expensive variant of these fans also have specific features that can be easily observed and distinguished even in normal, “home” use. Read more “Arctic P14 PWM PST CO or ball vs. fluid bearings” »


Arctic P14 PWM PST: Unbeatable in its segment

What is fascinating about the Arctic P14 is the particularly high contrast of price to (cooling) efficiency. These fans are among the cheapest, while at the same time achieving top results with respect to all 140 mm fans in terms of airflow per unit of noise. And that’s even through obstacles. Arctic has made almost the maximum out of the funds available to produce the fan, and it is definitely worth it. Read more “Arctic P14 PWM PST: Unbeatable in its segment” »


Comments (43) Add comment

  1. Will be reviewing Nidec SERVO’s GentleTyphoon in the not too distant future? Looking forward to the comparison with the A12x25.

          1. Personally, I think that maybe pushing is more often seen in the actual use of hexagonal grille(such as the exhaust position at the back of the case)?

            Thank you for detailed and useful reviews!

            1. From the perspective you’re describing it from, a grille behind the fan seems like a better/more practical option, but it’s a bit more complicated. I should note that we have spent an extremely large amount of time developing the methodology for fan testing. This is several years of work, where a team of physicists (and among them some who have been active in aircraft development in the past) and I analyzed the sub-issues in great detail (especially around the actual design of the test tunnel in relation to the fans being tested, and to make the results naturally as close as possible to real life), and it was in the matter of the grille installation that it was quite clear.

              It is important to note that its position alone has almost no effect on the performance characteristics of the fans, and it always throttles the airflow at the same intensity. With dust filters, the opposite situation (i.e. placing the obstacle in front of the rotor) again makes sense from a “PC” point of view, but it really doesn’t matter and the airflow restriction is at a comparable level in both situations. In any case, to achieve the same conditions (front throttling), we chose the same orientation to correlate as closely as possible the degree of the closed cross section (by an obstacle) with airflow loss. These are all complex matters and neglected topics for the commercial notion of tests, which nobody has ever properly addressed.

              So we will have a lot of work to do in the future to shed more light on this fan issue. Gradually, in some reasonable sequence (to make sense, we can’t start from the end) we will publish various, say educational articles on these and related topics around aerodynamics in the context of application in PCs. I understand, and I am not surprised, that many PC users find these articles of ours indigestible, but we will do everything in our power to reverse this unfavorable situation. 🙂

                1. We’re very thankful for your constructive thinking. Keep it up, you’re an endangered species. 🙂

                  1. Another reason why the A12x25 current efficiency is not particularly excellent should be the choice to increase the air gap between the permanent magnet and the stator of the brushless motor in order to reduce vibration and noise. This is one of the common and effective means of vibration and noise reduction for single-phase DC brushless fans (e.g., many of Arctic’s fans), and the deterioration of commutation losses is acceptable for such low RPM fans.(However, for high RPM fans in industrial applications, high efficiency and reliability are often preferred)

              1. I don’t find your articles indigestible, but I can certainly understand why they would be confusing for the average end-user. I think that “dumbing it down” might help, in the sense that you offer a basic presentation for those who don’t want the technical stuff, and have a sortable table so they can choose a fan based on their needs and preferences (although I hate the idea of a table being sorted on LEDs!). I wonder if you do short YT videos to present your findings, which can give you a bigger audience and help pull people here for more details.

                Thank you so much for your hard work, and I hope you’ll be able to advise me!

                1. We are not interested in videos and especially not in YouTube, as far as some knowledge from tests is concerned. We value our work, the sweat and blood, too much for that. There are some honest, well-meaning creators on this platform (like you), but there are also many speculators who use their great influence to egotistically manipulate the public debate. And we don’t want to be in the same boat with such people.

                  Moreover, things on YT can only be “successful” if they are simple enough and at the same time evoke strong emotions. And neither of these applies to fans. Fans are a very complex subject to understand and any sensationalism around them is always nonsense that only works on people who have little idea of how a fan works.

                  Another thing is that with videos there is no way to clearly present the amount of data we consider relevant for a full evaluation. Someone may only be interested in 5% of the stuff, but they can pick out exactly what they are interested in right away and not have to wait for the author of the video to elaborate on things that the viewer is not interested in. This is the great advantage of text articles, as long as they are clearly structured. For the presentation of results to the extent that we do, I consider the format of the videos to be completely inappropriate. And oversimplification in topics that require a certain level of complexity to make the conclusions accurate is a road to hell (that’s just more confusing, lack of detail, too few angles on things, and cherry-picking).

                  Sure, they are more difficult articles to understand, but they are presented in exactly the way we identify with. With the amount of time we put into it, it would be impossible to motivate ourselves with work that we ourselves would not be satisfied with. So we consider the text form full of graphs, charts, technical analyses and explanations of various phenomena to be the most suitable platform. We don’t have to impress everyone at all costs. Things that too many people revolve around usually have very low moral value.

                  For a better understanding of many things by regular users, we will be publishing a series of articles from next month on various topics that we consider important to dissect in a detailed way and present in a popular way. Of course, we want to gain new readers and ignite for this topic also those who have not been interested in it so far. But again, we will never sink to a level that is against our worldview. Professional publications (and we don’t want to and won’t do anything else) have to meet a certain standard, period! 🙂

  2. Perhaps in the side of the frame to test the noise sound pressure value, the phase difference between the two ends of the fan blades will interfere with the sound pressure test results at different distances(for example, two fans will be at different distances on the side of the frame to produce different sound pressure comparisons).

    This is one of the best fan reviews I’ve ever seen, and it’s been very helpful, thank you all for your work!

    1. Thanks, fixed. There was a wrong path to the charts in the source code (with the wrong month), so they weren’t showing up. But everything is fine now.

  3. I wish you (or someone else, just a single publisher would be enough) would make measurements in the usable speed range, I don’t have tools to measure the acceptable noise but I can tell you , nearly 800RPM is way too much for this fan (and obviously even more so for worse ones)

    while it’s clearly the best fan on the market it’s not as magical as many claim and it becomes too loud at around 600RPM already
    the rule is simple: if you can hear it (from the reasonable range, PC under the desk and so on) it’s broken so your lowest measurement is already exceeding the usable range

    everything else about your procedure is amazing, unfortunately all measurements are made at above the acceptable noise level

    1. Thank you for your comment (and the praise at the end of it), but I have to express some disagreement. You seem to be talking about normalizing fans by speed level, which I am not too happy about. Yes, the NF-A12x25 has 787 rpm in the quietest mode of the fixed noise levels, but many other models have speeds below “your” 600 rpm. For some, it’s below 400 rpm.

      Fan speed doesn’t say anything about noise level. A more interesting indicator is what the fan speed is at the same noise level. Although even this is rather useless information, because it lacks any intersection with performance characteristics (with airflow). We know these with the NF-A12x25 (they are top notch) and the fact that some of the highest speeds are achieved is indicative of the qualities of the elements that suppress the high noise levels at critical/dominant frequencies of the noise spectrum compared to other fans. Indeed, our “31 dBA” mode is extremely quiet, and the amplification collar we use is needed to distinguish it, in addition to the small spacing. The airflow of weaker, less efficient fans in this mode is often very low, at the limit of measurability.

      TL;DR: Not every 600 rpm is the same, what matters is the noise level in dBA, which is mainly shaped by how effectively the fan manufacturer was able to suppress the noise intensity at the dominant frequencies of the noise spectrum.

      1. From what I understand, zakius probably is highly sensitive to noise and owns an A12x25 themselves. With their excellent hearing, they consider even 600 RPM on the A12x25 too loud for their taste. So basically they are asking you to offer tests at even lower noise level thresholds.

        1. If this is the case, then of course we can include this in the additional NF-A12x25 PWM tests, where we will investigate the effect of the colour/strength of the material on the aerodynamic properties. When zaikus writes out what specific speeds he’s interested in, I can run that fan across all modes at those speeds. But since it will already be below the resolution range of the noise meter, it will be without dBA measurements. Anyway, it might be interesting to see how many aerodynamically less efficient fans the NF-A12x25 can beat at a clearly lower noise level.

          1. It’s a shame you are unable to compare other fans at the same noise levels, but I understand it requires scientific grade equipment that costs a lot more than the meters used by basically everyone and the demand for such tests is negligible. And exactly because such meters are hard and expensive to get I can’t offer any numbers for the noise levels and offered the speed of tested A12x25 that in my personal environment is still satisfying my acoustic needs. I can say it isn’t moving much air and definitely wouldn’t try that with an LC radiator or densely packed fins stack but it certainly is improving performance of NH-P1 by gently guiding the air through it.

            In general the issue I often have with PC cooling components reviews comes from people used to sitting with AC on and if they can’t hear the PC over it they claim it’s silent while for me it often means a lot of constant noise, for example I own Dark Rock 4 and Ninja 5, both known to be “extremely quiet”, some people even claiming that Ninja 5 could’ve used the regular variant of Kaze Flex instead of the 800RPM one since it’s sacrificing performance for “no benefit in acoustics” and are recommending Fuma 2 instead, and I personally not only couldn’t stand these fans at 800RPM, I still found them obnoxious at the lowest speed they actually ran without issues.

            Additionally it’s nice seeing you were able to pick up some parts branded SilentiumPC and show how they truly perform, people here are pretty much in love with the brand as they offer decent performance per price, though when it comes to acoustics there’s an enormous gap to close, even the latest Fluctus is severely underperforming. I wonder how would Arctic P12 do as it’s another cheap fan people love to praise. And on the higher end of things eLoop B12 that in my experience does pretty nicely when it comes to blades noise but the motor is really whiney

            1. The words about “unscientific” equipment, “like basically everyone else” hurt a bit. 🙂 The development around all the stuff was really crazy to make us happy with the result ourselves. But I guess I can rise above that and understand that from the point of view of anyone who can’t see what’s behind it (and hasn’t been with us through the x years of torture, tuning, studying, first, second and third prototype test tunnels and its xy revisions), the more impressive thing is the expensive machine that’s listed for xyz thousand USD, but that’s really not what it’s about. Intellectual capital is always the most valuable in everything.

              I can think of one example from a slightly different world: Even the most expensive soccer cleats won’t make a below-average soccer player world-class, just like cheap soccer cleats won’t make a world-class soccer player average. This may also be the reason why Gamers Nexus hasn’t yet decided to run fan tests, even though they’ve invested a fortune in measuring equipment. The difficulty for them is that this is an extremely difficult subject requiring in-depth knowledge of various areas of physics for proper measurements (and interpretation of those measurements, evaluations) and by not going through the extreme work of developing their own measurement setups (tunnel, acoustic security), they obviously don’t feel the ground under their feet is too solid.

              But back to the NF-A12x25 PWM and low speeds below 600 rpm. Even if we could distinguish this level of noise, we are already below the start-up speeds of the vast majority of other fans. This means that even if we could set the NF-A12x25 to some controlled, extremely low noise level, there would be nothing to compare the fan to. Something like the a Phanteks T30 or the BeQuiet! Silent Wings Pro 4 (i.e. the biggest competitors) wouldn’t be able to get down there, after all. Other fans, while aerodynamically quiet, will again be limited by “too noisy” bearing friction, and in the end we’ll really only have the results of the NF-A12x25 in that mode, and maybe two or three more fans that won’t even be of interest (because the low noise will be achieved by them at the expense of lower performance/airflow).

              1. I’m sorry if that sounded like that, I meant the sound meter alone since you stated it’s already at its limits, and I agree that most of fans wouldn’t even meet the required noise levels anyway. Unfortunately there’s no place to compare fans (and other cooling equipment, I’d love to see AIO and tower coolers with different fin stack structures thrown into such a strict noise measurement procedure) when you are targeting so strict acoustic norms and while it’s very likely A12x25 is the best one in that range there’s no evidence for that and when finally new model emerges even from Noctua we can’t automatically assume it will be better in that aspect.

                That’s just my (unrealistic) dream to be able to reliably pick the best one, and as stated it not only needs proper anechoic chamber since it’s too easy to disturb the measurements when we’re aiming so low but also an extremely sensitive and precise meter like SVAN 979 that costs easily 10x the price of already high quality professional meters.

                1. Sure, I understand. Of course, when it comes to details, it is possible to go into complete ridiculousness and things can always be improved in some way. And as our abilities change, we will try to do that over the years. For now we can only do what the extremely low (basically no :)) valuation of this type of work allows us to do. When companies and people start to value professional approach more (instead of the consumptive one), then of course we can tune such details as even better noise meter, even better acoustic conditions and so on. But even at this stage, we have overshot beyond what 99.999 % of users expect/need. We don’t want anything in the test procedures that doesn’t paint a complete picture of the fans in all aspects. Clearly, an even lower fixed noise level for the most sensitive users… I firmly believe that we’ll get there one day. If you support us with comments, even of this type (and show the world that there is interest in such measurements), now and then you conveniently share some of our tests somewhere, then even your “unrealistic dreams” and serious dreams can have a sharp outline. I can’t tell you how much I would like to realize them. ?

                  Unfortunately there’s no place to compare fans (and other cooling equipment, I’d love to see AIO and tower coolers with different fin stack structures thrown into such a strict noise measurement procedure)” this is really impossible. Even if you test all fans on all types of heatsinks on all processors at all thermal loads, the ratios will be broken by the different settings of system fans in the case (not to mention how restrictive the resistance of the environment is to the fan, which depends not only on the intensity of the airflow in the case, but mainly on whether the fans create overpressure, underpressure, or equal pressure in the case). How to partially limit the dependence of the radiator structure (and that it is damn necessary!) on the replicability of the intended results, we have discussed in this article. In hindsight, even based on the feedback from the readers, I recognized that the interpretation is rather vague, brief and, in short, hard to understand for someone who hasn’t studied the dependencies (and can’t see deeper into it). So we will deal with these topics later in much more detail, with simpler interpretation, examples, tests. A sketch to make the unclear things clear. I just need enough time (and a fresh mind), which I don’t have for these detailed analyses, and because of the time demand on the tests I can’t keep up with them permanently, and I’m pretty exhausted from everything and I’m glad I can go to sleep at least for my four hours.

                  PS: It’s a bit of a pity that these things are not dealt with in more detail by big manufacturers with big budgets, with a big team of people, and they are satisfied with just misleading P/Q curves. But it’s hard to expect any significant progress there, when nobody is willing to include even the noise level of the whole cooler in the parameters and the standard in the specifications is only the noise level of the fans themselves without the sound added by the microtubulations and the resistance of the heatsink. ?

                  1. In the perfect world the manufacturer would just publish the full speed to noise graph at the very least and that would save a lot of hassle for reviewers, and I’m pretty sure companies already have equipment capable of measuring their own products at least. As long as that data would be reliable and true. Though on the other hand there’s much more to fans, some of them behave poorly when mounted as pull, some may misbehave when running on different voltage, even if the speed stays the same, and so on, it’s impossible to test *everything*.

                    When speaking about AIO I meant that there’s so much more than just fans that can make noise there, obviously there’s a pump, but if we’re aiming low enough all the hydraulic parts can make a difference too.

                    Regarding your last point I’m disappointed that Noctua does as you stated, listing the fan specs instead of the whole assembly, but beQuiet! actually does it right, at least for coolers using fans available separately, in these cases they list the assembly noise levels and beneath that the fans noise levels, for coolers using fans available only with them there are only the whole assembly numbers available. Obviously that may not be true for all the products in their offer, but that’s what I managed to gather by glancing over it

                    1. BeQuiet! really takes into account the effect of the heatsink when specifying the noise level. They are also the only ones in these topics (noise perception and the like) to create educational materials for the masses. Noctua does go into detail in describing the various elements (although it’s hard to say how much the general user understands them), but these always relate in some way to the technologies they use in their cooling systems. They don’t go into general terms that don’t directly promote their cooling stuff. But unlike many companies, they at least clearly separate heatsink and fan specifications when it comes to coolers. Some (probably most?) companies mix everything together and it looks like the fan noise corresponds to the overall noise of the cooling system/cooler (heatsink + fan(s)).

                      Companies that have the financial resources don’t want to invest (from a profit point of view, i guess it’s not worth it) even in such basic things and we would, of course, like to use the best noise meter with the lowest range in the quietest place in the world, but that is, of course, utopia. Doing things honestly in hardware testing is a huge “handicap”, which is why nobody does it anymore and why mainstream hardware tests look the way they do. People want to “be entertained” first of all (that’s why they are tabloidized) and ideally at someone’s expense, and then there is not much room left for those who want to push the current boundaries of knowledge. Many companies even frown upon this, because precise things often clash with the marketing ones which sell. 🙂

          2. Measurements without any relation to noise make very little sense in my opinion. I’m an extremely noise-sensitive person as well. In the evening, I can hear an Arctic F9 fan at 600 RPM from 3 meters even though it’s more silent than Arctic P12 at the same RPM. I have a full mesh case and I’m changing it to Nanoxia Deep Silence 8 Pro, mainly to combat coil whine that is far more annoying than any moderate fan noise.

  4. Impressive that NF-A12x25 managed to beat the newly released Silent Wings Pro 4. Arctic P12 next? A word of warning – getting a sample without fan-frame friction and motor problems is 50/50 at best. I’m not amazed by its performance, but it would be great to see how this popular fan does against much more expensive competition.

  5. Hey there, I loved this super in depth review and it’s narrowed it down a lot for me with my choices for my system. I have an interesting setup and wanted some input for it, if you could advise me with any tips. I have a Coolermaster NR200P case with the inverted mod, but basically just have a 240mm rad for the AIO GPU at the bottom of the case, and a 280mm rad for the AIO CPU at the top of the case. Not much room for fans on the side panels as far as I know (the GPU blocks one side panel, and the other one I could squeeze a fan in, but the airflow is obstructed by the components in the middle of the case). Do you know what the optimal fan configuration is? I’m thinking I should run intake on the CPU rad fans on the top and exhaust for the bottom GPU rad fans, what do you think? And thanks a bunch for the reviews again.

    1. Thank you for the question.

      Without knowing the TDP of the components used and the restrictiveness of the filters, grilles and overall design of the case, it is quite difficult to give specific tips on fan settings. However, we can at least start from some general rules, which include the fact that from the point of view of achieving the best cooling performance/noise ratio, it is more effective to set the speed of the fans so that underpressure is created in the case. We also covered this in the past as part of the thematic tests in the article “System cooling in the spotlight. What to increase/decrease?“. Fans on the exhaust are more important from a cooling point of view. Cooling with two fans at the exhaust would not be optimal, but it would be much more efficient than using two fans at the intake (and leaving only the perforation at the exhaust, but nothing active…). Intake fans, of course, make sense, especially in configurations with powerful graphics cards, so your reasoning is right – using two for intake and two for exhaust. In line with what I wrote above, to achieve the lowest possible noise at the highest possible cooling performance, I recommend lower airflow (lower speed) at the intake and higher at the exhaust.

      I recommend the Noctua NA-IS1 frame for intake fans, so you can reduce the intensity of microturbulence on the sharp edges of the bottom grille of the Cooler Master NR200P, which will mainly reduce the noise, but at the same time for laminar / smoother airflow will increase the airflow a little. Without NA-IS1 (or a similar frame), it could happen that even though the intake fans will be slower, with lower airflow, they will be noisier than the exhaust fans.

  6. Hello,

    Thank you very much for such detailed review and all the efforts you put in your work !
    Since I discovered your website I learned a tons on system cooling.

    By the way, do you think you could make the same review for the NF-S12A PWM, according to noctua it’s supposed to be used as a case fan (and if I understood correctly you used it as an exhaust fan), so I guess it should beat the NF-A12x25 in the “hexagonal grill” scenario (for both pressure and airflow),but I’d like to know for sure.

    I can’t wait to read more fan tests from you

    1. I did airflow test for NF-A12X25, NF-F12, NF-P12 and NF-S12A on 30T radiator push setting. Note that a thinner radiator has slightly more impedance than a hexagonal grill.

      For noise-normalized airflow, NF-A12X25 had 20~25%p higher airflow than NF-S12A. NF-F12 and NF-P12 showed similar result(but NF-P12 was a bit lower due to resonant noise at some speeds.) and NF-A12X25 was 10~15%p higher than them. Therefore, the efficiency order between 4 Noctua fans is: NF-A12X25 > NF-F12 ≥ NF-P12 > NF-S12A

      Even a moderate impedance like a thinner radiator can work as a huge penalty against NF-S12A. A hexagonal grill has a bit lower impedance than a thinner radiator, so the result for a grill with grids may be similar. The only scenario that NF-S12A can shine is an absence of any obstacle. Of course, this is an impossible condition for PC cooling!

      My test is not precise as HWCooling’s excellent benchmark, but I hope this will be helpful for you: (Please use tranlator to overcome a language barrier.)

      1. The results on hexagonal grilles are very difficult to generalize because of their large structural differences. The size of the cut-outs is similar, but the spacing between them is different, which significantly changes the number of edges, but especially the degree of opening. Some hexagonal grilles are open up to 80 %, others only 30 %. Ours is perforated at 50% (a golden middle way – although in practice the grilles are often more open and with lower restriction. But at least individual fans have their strengths or weaknesses more pronounced in such applications, on the grille). Anyway, it would require an extensive test of as many grilles as possible… not only hexagonal, but with the different shapes of cut-outs that are used. And all types in combination from large opening rate to small. Best to test it with at least twenty fans gradually from very low to higher speeds, haha.

        1. I overlooked the parameters of grill. (grid thickness, total opened area, area of unit cell and etc..) Maybe there are some conditions that NF-S12A can over-perform NF-A12X25, but I still think it would be difficult for NF-S12A to beat NF-F12 or NF-A12X25 under most conditions of grill due to its relatively unefficient design in obstacles.

          Anyway, it will be very long journey to test those fans on many types of grills! I look forward to seeing it posted someday.

    2. Sure, I distinguish between your tests and QM오즈’s tests. I came across them about half a year ago when you confronted your NF-A12x25 results with ours.

      Cheaper noise meters, but I don’t want to pausalize it, can distort frequency peaks more, according to my experience. One way or the other. That is, they either inflate them or undercut them. Thus, fans like Arctic P12 or BeQuiet Silent Wings Pro 4 can be over or under-rated. The calibrated Reed R8080, which we use, scales nicely with the calibrated MiniDSP UMIK-1 microphone. It is very accurate except for very high frequencies (above 10 kHz), which the sound of the fans does not contain. So yes, your reasoning is correct, in fan tests with normalized noise levels, the order of the fans can also be determined by the characteristics of the noise meter. But that doesn’t change the fact that your work on fan tests is very useful and you are definitely in the mix of the most accurate publications on computer fans on the internet.

      These things around sound recording are extremely difficult to implement. To measure “the same thing” twice (with negligible variation) is acoustic madness. We had to tune the acoustic panel set for a very long time so that, across repeated measurements at the same fan setting, it would produce consistent results at all frequencies.

      1. Thank you for the important insight about noise measurements. Now I understand why the efficiency orders between the fans can be affected by the noise meter itself. One of the reason of slightly different result depends on how severe the frequency distortion in my noise meter is. Next time I post other test, I need to noitice this to my readers to prevent misunderstanding or confusion.

        But it is also the fact that currently It is impossible to build more precise noise bench for me due to limited finance and space. So I decided to focus on the diversity(especially, by fan size) of fan tests with ‘compromised’ precision.

    3. By the end of the year we will surely add tests of Noctua S fans. The NF-S12A will be preceded by the older NF-S12B design, so that it’s complete.

      The graphs with P/Q curves are interesting until one realizes that “PC case” is a very broad, actually non-specific term (a fan in a case can be in an almost completely non-restrictive environment, but also on a grille with a dust filter) and that they do not in any way take into account the noise level. For example, a plastic filter will increase the fan noise (by increasing the noise at higher frequencies 5–7,5 kHz characterized by hissing) more than a radiator.

  7. Usually cases have both a hexagon grill and a nylon filter, like the main bottom intake of my Silverstone FT02.

    Maybe you could do a test that combines both like this

    1. Thank you for your message.

      Yes, you are right. In practice, there are a million plus situations. For obvious reasons we cannot have all the combinations in the standard methodology, so we have only basic situations, which can be used to predict the behaviour in a specific situation. A fan that significantly loses airflow on a grille and suffers with a nylon filter will be below average in the combination of these two obstacles as well.

      In some cases it may happen that the filter in symbiosis with the grille can reduce the tonal peaks by changing the characteristic of the airflow, but these are matters for separate analyses focused on a specific case. If in the standard methodology we had that one extra obstacle (filter + grille), then the realization in practice can be done, in a combination of different subtleties of filters and grilles (and for them the shape of the cut also plays a role in the result) in a thousand ways. But we will break everything down to the smallest details as time goes by. But this will be done in separate articles, for which we will select only a few fans. Stay tuned. 🙂

    1. I would take what Igor publishes with a very, very large grain of salt. It starts from the basics. It is physically impossible for spectrograms to differ so diametrically with a fan that has the same geometry. It may be that there will be different tonal peaks (due to the eventual different vibrations caused by different material strength), but the noise floor will not be shifted across all frequencies. This is probably increased by the electromagnetic noise, which was different in the past than it is now. Just use a different device to which the same microphone is connected. This is why battery-powered noise meters are always more reliable, although they do not provide such a detailed spectral analysis.

      Wallosek unfortunately focuses on sensationalism at any cost, similar to Gamers Nexus. That’s what people click on, that’s what makes them money. But it’s irresponsible and I have to speak out now and at any time in the future. Igor, moreover, has completely lost his judgement, when he discredits, for example, the Longwin LW-9266 without any foundation. And in combination with how inconsistent he is at first sight, it is really courageous. Note for example the unreasonably rough fan mounting, which would be fine on a henhouse, but for fan tests where extreme sensitivity is required, it’s a real problem. Especially when the author claims to measure frequency analysis of sound and this faulty approach alone distorts it. Each pad intervenes in front of the fan to a different degree and thus the effect on the noise level or the height of the tonal peaks of the resonant frequencies is always different. Now he cares a little more about the mounting, but even so, there can be no word about satisfactory repeatability of measurements with regard to comparing the properties from piece to piece. He doesn’t have it under control at all. He does work with a lot of technical details, but these are irrelevant for the final result. It’s a similar approach to the way Gamers Nexus works. Focusing on the unimportant things and the important things… somehow didn’t get through. I consider this very dangerous, because it makes the average reader feel excited and they can be manipulated, but I don’t really buy it, I just smile or sometimes even get a little bit angry, because it unfairly distorts people’s buying behaviour… and Igor can rant as much as he wants, that he meets with a negative reaction (not only from my side). If he approached these things about tests with a bit more humility and could appreciate someone else besides himself, this reaction would also look quite different, milder. But it cannot be ruled out that Igor is quite lacking in self-reflection.

      We did not come into contact with a NF-A12x25 sample with a production year of 2023, but we cannot measure the differences between early samples (2018) and those from 2022 even using our methodology. I wanted to compare the variants with the beige-brown ones, but it probably won’t happen (we don’t have time for that and we will leave such a comparison for the new generation of 140 mm fans). If the new NF-A12x25 were really noisier, it would have to mean that either Noctua went softer on manufacturing tolerances or they use a more flexible material for the rotor, and both seem extremely unlikely to me. After all, it’s pretty irresponsible to base a test like this on a single sample.

      1. Thanks for the quick response and insights, I am glad we arrived at the same conclusions.

        I also looked at the graphs and even listened to their recording, and found it really strange that the 2020 sample seems to merely have a raised noise floor instead of change in noise profile like you’d expect if the fans are really different. I was leaning on measurement error, but they claim that their results are repeatable so I doubted myself a bit (the doubt has been cleared now 🙂). Honestly their whole measurement setup are full of questionable design choices with no good rationale to back them up, from mounting, to airflow straightener, to noise dampening, and to the measurement or airflow and noise… And they just handwave the details by saying there’s “foreign know-how” and they would be “too far for the normal consumer.

        I hope you’re wrong with Gamers Nexus, they recently made a anechoic chamber and is trying to do frequency analysis. They *seem* to be moving in the right direction, though I am being very cautious as they have poor track record in doing proper noise analysis.

Leave a Reply

Your email address will not be published. Required fields are marked *