Testing Fan Differences
And that’s why we’re here today: We’re benchmarking Noctua’s NF-A14 fans from both China and Taiwan, having sourced several of our own, official samples from Noctua, and the samples of the original Reddit user. Testing primarily consists of airflow testing using an anemometer and a hotwire (+/-2% accuracy), noise testing (dBA), and PWM to RPM response signal testing. Thermals won’t be covered here because, frankly, it’s irrelevant: Any thermal differences detected by these fans would either be sufficiently explained by FPM measurements or by normal test variance. This type of thermal testing has far too many variables introduced by the host system and its software, and so we’ve eliminated it for lack of adequate test resolution. FPM, RPM, and dBA are all we need. Frequency spectrum analysis would be nice, but we didn’t go that far for today – though we do have some comments on subjective pitch output.
An NZXT Kraken X42 (140mm) radiator was used as a flow straightener and to provide a common object where someone might need a static pressure-optimized fan. Data collected over a few minutes to allow local eddy currents to form and dispel.
As for Noctua’s part in all this, we got into contact with the company (for the first time ever, for the record) around when the first post was made. Noctua informed us that the company has made fans in both China and Taiwan for a long time now, but that they normally do not produce the same fan at both factories. In order to keep up with demand, Noctua made a new mold for the China factory (which was already working with the company) to start producing NF-A14 and other fans. Both the NF-F12 and NF-A14 have been produced primarily in China over the last few years, so anyone selling Taiwan-made variants likely is just moving old inventory. This limits the relevance of color-matching for the fans, as any new buyers will mostly receive fans from China.
Regardless, the question was whether the origin of the fans actually impacted quality or performance. Visual differences are present in the color (slightly different beige, slightly different brown – the China model tends to be darker in fan blade coloration). Other minor differences are visible in Noctua’s marketing items, like the “flow accelerators” and “stepped inlet design.” Little triangles in the frame are slightly ovaled-out in one fan and have sharper lines in the other. This seems inconsequential, but we’ll find out below.
PWM Fan Speeds (And Why PWM is Unreliable for These Tests)
Let’s first talk PWM to set the stage. PWM fan control varies from fan-to-fan, and fan manufacturing differences mean that your top-end and low-end RPM is going to change between units. The same is true for water pumps, and this is true across the industry. In general, PWM variance can be up to 15% between fans, so you really want to fix the fan speeds if testing differences in frames and blades, since otherwise you’re just testing normal PWM variance. Noctua noted to us that PWM to RPM response typically varies by 1-4% in their products, with a tolerance of 10% in PWM range – pretty standard.
If just testing with PWM fan speed percentages, the numbers are going to be all over the place. It’s not a reliable way to do things; regardless, we’ll start there, then move to testing that shows performance with fixed fan RPMs to illustrate differences in design. We are also looking at PWM range to see if the China-made or Taiwan-made fans are consistently lower or higher in RPM.
PWM to RPM Response – Fan Speed Range
Here’s a line plot showing RPM on the vertical axis versus PWM signal on the horizontal axis. The fans are all close to one another here, with normal variances dictating the difference. Averaged across all tested fans and all PWM ranges, we found an average difference of 2.5% in RPM to PWM response between fans, with no particular favor shown toward China or Taiwan manufacturing. Note again that we are limited by our sample size, so we can’t make sweeping statements that this will always be true – but it is true for our tests. We see no significant difference between region of origin. And again, variance in PWM signal to response is normal, anyway.
Adjusting PWM speed by +/-1% often will allow RPMs to match, as we’ll do in a moment.
|PWM 100%||PWM 90%||PWM 80%||PWM 70%||PWM 60%||PWM 50%||PWM 40%||PWM 30%||PWM 20%|
|AF-14 Taiwan #1||1513.1||1404.4||1262.7||1141.5||997.1||845.6||683.9||528.8||340.7|
|AF-14 China #1||1500.7||1397.2||1285.0||1149.2||1006.3||864.6||692.3||536.6||334.0|
|AF-14 Taiwan #2||1485.2||1371.3||1249.7||1121.3||983.7||837.8||674.3||521.3||335.7|
|AF-14 China #2||1514.1||1400.3||1271.5||1145.4||987.5||839.6||678.5||524.7||335.8|
|AF-14 Taiwan #3||1500.8||1398.0||1285.0||1145.0||1004.0||860.4||688.0||530.0||330.0|
|AF-14 China #3||1501.6||1396.2||1268.8||1167.7||1005.5||854.4||703.6||545.1||343.1|
Here’s a look at a table of just some of the RPMs at each PWM range, just to illustrate this point numerically. This variance is normal between all fans, and is not a product of being made in China versus Taiwan.
PWM vs. Noise Response
Continuing on with the PWM point for a minute, here’s a chart showing noise levels at the varying PWM ranges, keeping in mind that fan RPM changes between fans at each percentage. At 100% PWM, our dBA range is 42.8 to 43.6 – not noticeable, and more or less within our test accuracy. 90% posts a range of 40.6 to 41.4dBA. 80% posts 38.2 to 39.1dBA, with no particular benefit on the China or Taiwan-made fans – that said, a single fan from China, sent by the reddit original poster, did generally perform about 1dBA louder than the quietest fan.
This continues scaling down and the fans get closer as we go, with our ability to test halting at around the 27-28dBA area.
Let’s fix the RPM levels to get a better idea for performance.
Fixed RPM vs. Noise Response
Manually setting fan speed in BIOS to match 1400RPM or 1000RPM, we end up with the numbers below. At 1400RPM, every fan – regardless of country of origin – is within fractions of a decibel of 41dBA. The differences here are imperceptible to the human ear. You would not be able to identify one to the next in a blind test. Although we did not produce a frequency spectrum plot, we did not notice significant differences in fan frequency output.
At 1000RPM, every fan is within +/- 0.5dBA of 33dBA, making them again effectively identical in noise levels. We did not subjectively notice differences in frequency spread.
PWM vs. Linear Feet Per Minute Airflow
Moving on next to linear feet per minute testing with an anemometer logging over a period of several minutes, using the X42 radiator as a flow straightener, here’s a look at linear FPM response to PWM signal. Again, keep in mind that PWM signal doesn’t mean much – the fan RPMs are not the same, so performance is different as a result of RPM changing, not necessarily as a result of fan body manufacturing quality. We’ll look at that aspect next. Linear feet per minute will be abbreviated as FPM for this section.
FPM posts a difference of +/-2% from 20% to 100% PWM, with no favor toward China or Taiwan fans. Again, our difference here stems from PWM to RPM response variance, which is within normal manufacturing parameters.
We’re averaging around 515-520FPM at 100% duty cycle, about 460-477FPM 90%, and so on down the line. All results are fairly tight together.
Fixed RPM vs. Linear Feet Per Minute Airflow
At a fixed fan speeds of 1400RPM, the numbers are largely the same across the board. The only stand-out is the first Taiwan-made fan, which technically plots a 6FPM lower result than average. This is within our test margins. Without better equipment, we can’t dig deeper than this. As far as our testing is concerned, these are all effectively equal and will produce indistinguishably different thermal results given the proximity in FPM. Differences in thermal testing would stem more from all the variables in a CPU test than from the fans, it seems.
At 1000RPM, it’s more of the same story. We’re close enough in performance to write this off as more or less equal. Greater test resolution and controls would be required to plot differences reliably, at which point we’re exiting what the user can detect, anyway. Speaking regionally, there’s no pattern between the two sources.
Conclusion: The Internet Outrage Engine
This whole debacle started off with good intentions: A user reported product-level concerns that prompted further investigation, as it’s always possible that media samples either don’t exhibit issues or are limited in quantity. This is good, particularly when the reporting is done with the consideration that there’s always room for a one-off defect, user error, shipping damage, or some other one-time event; unfortunately, like most curiosities and slight offenses posted on the internet, the trouble cascaded out of control until people were threatening to stop buying Noctua ever again. It’s an interesting, insane flip that internet outrage culture causes: in the blink of an eye, we go from, “I trust this product and have liked it for years, have recommended it, and presently use it for critical systems” to “I’m never buying this again, I can’t believe they did this, and I’ll cost them so many sales they won’t believe it.” That brief injection of anger satisfies the itch to hoist someone to the gallows – this time, it was Noctua.
People are binary in this way, if our CPU and GPU coverage has taught us anything. There’s little room in internet comments for a saner middle-ground between raving praise and seething rage. Brand loyalty rules over thought, and the last week demonstrated how impossibly high standards and unrepentent loyalty turn a small fault into a warfront.
We can’t speak for every single Noctua fan that came off the line, but we can speak for our sample size – and it is a decent sample size, all things considered. It’s not like we’re a validation house that was contracted to test hundreds of these, after all. Between our eight presently tested samples, we cannot see any meaningful difference using our measurements. Visible differences, yes, but that’s about it. This sample size includes fans from the original Reddit post, sent by OP for t esting. Noctua might deserve a bop on the nose for the color change, though that’s an awfully small thing to boycott over – pigment and plastic coloring is going to change by region. It’s difficult to color match cross-region, and this looks like an instance where Noctua either didn’t see the issue, didn’t think it was an issue, or didn’t have time to respond to the issue. And “issue,” all things considered, is a heavy word for “beige” vs. “slightly more beige.”
Noctua is fighting its own reputation for quality and precision. Because the company has built the brand around these two attributes, a single slight – big or small – will immediately invoke polarized responses from the community. In this instance, Noctua was chastised more than deserved. We could not find, using our samples, any meaningful difference between the China- and Taiwan-made fans. This is an instance of something getting blown out of proportion by Reddit, which seems quite frequent.
Companies change supply and factories all the time. People notice differences all the time, too. The only change here is that Noctua was noticed (doing something mundane and uneventful) and, for whatever reason, the PC hardware community latched onto it for lack of things at which to direct pitchforks. Cooler Master changes the fan used with its oft recommended Hyper 212 – and those differences are far greater than these. NAND supply changes. DRAM supply changes. Even the switches used in mice change. Ultimately, all that matters is that the product matches the spec that it has always advertised; if it does, then no big deal. This is why vendors often don’t disclose suppliers for, as an example, mouse switches: They know we can figure it out, but they also know that Omron might not fill demand forever, and so leave room in marketing language to swap suppliers. In this example, as long as both switches do the advertised X million clicks and feel the same, it really doesn’t matter.
Noctua isn’t even that extreme, here: The fans are slightly different in color. That’s about it, from what we can see in our testing.
Now again, that doesn’t speak to every single fan that came off the line. Maybe one is defective, or maybe a certain batch of fans had issues, from said batch we may not have obtained a fan. If that’s the case – if you find actual problems – please let us know so that we can look into it. Just remember, from above, that matching PWM signal is not a sufficient way to test differences.
Editorial, Testing: Steve Burke
Video: Andrew Coleman