Arctic Fan Roundup

Having tested four 120 mm fans from Arctic, I am in a good position to compare their models. I tested their PWM versions. I assume they perform similarly to their temperature controlled and voltage controlled variants. As you will see, the star of this article is the F12 Pro, but not in a good way. The models I have are

Prior to this comparison, I had measurements for 8 fans from various brands, and found that the Arctic BioniX model was the best performer. This motivated me to try other Arctic fans, especially the “Continuous Operation” fan, which is marketed as being more reliable, or at least designed for “continuous operation 24/7.” That is definitely a good design goal, but I find it odd that they don’t describe how reliable their other fans are. Are you supposed to only use Arctic fans for a only few hours at a time, and then they need a rest? Oddly, the CO model is cheaper than the BioniX model. Possibly the BioniX model is priced higher largely for its aesthetics.

Enough speculation! Here is my report, which you can generate at bench.pcfanreviews.com.

The Arctic F12 Pro is my new worst performer. Objectively it performs like the Nidec Servo Gentle Typhoon, which costs three times more, but subjectively the Gentle Typhoon is a heavier, more sturdy fan and so I would hope it is more reliable and has other good qualities.

The F12 Pro is a larger fan due to its depth of 38.5 mm rather than the standard 25 mm, and so it should perform better than average. Despite having 54% more volume for its bearing and fins, it performs worse than almost anything else.

I almost can’t enough bad things about the F12 Pro. Its performance isn’t its only problem. It only has mounting holes on one side, so if you want to reverse the airflow direction, you have to mount it on the outside of your case. It costs a dollar more than the F12. Its frame is flimsy and so if you hold it while it’s operating, the blades hit the frame. The blades cut me when I touched them. I was giving it a hug! Stupid fan. Noctua fans like hugs.

I’m not done. Its PWM control works backwards! At 10% duty cycle it barely spins, which is fine. At 7% it stops. From 6% to 0% it ramps back to full speed!

The other fans perform well. For quiet operation, the vanilla F12 appears to be the best of them, while for higher performance, the CO model is better. If I could afford it and didn’t want to deal with fan failures, I’d purchase the CO fan. Unfortunately I don’t have anything beyond Arctic’s marketing to base that opinion on. It is, however, a very good performer as far as its noise and airflow. Never buy the F12 Pro.

Fan Benchmark Interpretation

I published data I created with methods I’m happy with. My intent is to provide data that answers the question, what fan pushes the most air for the least amount of noise? I have begun to answer that question. Following is the graph from the bench for the Corsair ML210 and the Nidec Servo Gentle Typhoon.

Consider the blue, vertical line. The higher a fan is at 40 db(A), the more efficient it is. That is to say, it moves more air given the same noise level. In this case, the Corsair ML120 is superior.

Now consider the green, horizontal line. The farther to the right a fan is at 40 m^3/h, the less efficient it is. That is to say, it creates more noise to move the same amount of air.

The line segments also reflect the fan’s effective minimum and maximum airflow and noise. The ML120’s minimum airflow is slightly higher than the Gentle Typhoon, but it creates the airflow with much less noise. On the other hand, the ML120’s maximum airflow is much higher than the Gentle Typhoon’s, but it also creates much more noise. Furthermore, the ML120 is much more versatile, as its ranges both in noise and airflow are much greater than the Gentle Typhoon’s.

The lines aren’t perfect, but they are a decent general guide for how good a fan is. The following shows the actual data points used to create the line using least squares regression as implemented by Apache Commons Math. Each point is the average of three sound level measurements and about ten seconds of airflow.

Does this mean that the Gentle Typhoon is an inferior fan in every way? Not necessarily. At maximum load, it uses 22% less power than the ML120: 0.14 Amperes instead of 0.18. One fan might be more reliable than the other, or have benefits in a particular application due to its design.

Lastly, my measurements do not predict if a fan will behave poorly with the surface you are going to attach it to. One fan might have excellent efficiency according to my measurements and then have a terrible resonance with your PC case, whereas a fan that was worse for me might not cause the same resonance for you. For instance, the ML120 feels less balanced than the Gentle Typhoon. My working surface resonated with the ML120 very strongly compared to other fans. It motivated me to greatly increase the amount of foam and cloth I used between my sound meter and working surface. I don’t want to bias my results in favor of fans that just happen to not resonate for the table I’m working on.

Back to my original question. Of the four fans I currently have data for, which is the most efficient? Interestingly, it’s the Gentle Typhoon. It provides 1.51 m^3/h/db(A). The Noctua NF-A12x25 is close at 1.49 1.41 m^3/h/db(A). Given the limitations of my measurements, they’re the same. What sets them apart however isn’t the efficiency but the baseline amount of airflow. For the same amount of noise, the NF-A12x25 pushes 16% more air.

If you know what these fans look like you might predict they would behave similarly.

This leads to more questions, but I’m going to leave them unanswered:

  1. Is this fin style the most efficient for airflow per db(A)?
  2. Is fin design the most important factor in such efficiency?
  3. What is the difference between these two fans that makes their baseline airflows different?

Quiet vs “Loud” Computer Cases

I was a long believer in keeping my PC quiet. Part of that has been to always use a case marketed as silent or quiet. My first such case was an Antec Solo (Antec has removed it from their website), which I purchased in 2011. I found far too cramped, so a year later I got a Corsair D550, which I enjoyed until October of this year (2018), when I began to question my loyalty to silent PC cases.

Quiet PC cases have several drawbacks. They are heavier and more expensive. In my experience, they fit less hardware into the same amount of space. I suspect that the obstructions they add to airflow also reduce thermal performance. That is to say, under load, the temperature of your GPU and CPU will be higher, and so the components can overheat or wear out faster. Even worse, higher temperatures can cause fans to spin faster to compensate, causing additional noise. This negates the entire purpose of having a specially designed case for lower noise.

To see if the extra weight and money have really been worth it, I purchased a Thermaltake Core V21. This case is small yet roomy, and importantly contains nothing special to warrant being marketed as quiet. I took sound and thermal measurements while running torture tests on my CPU and GPU at the same time. The hardware is the same except the case, the motherboard, and the front case fans. The motherboard is the same brand, but I had to get a new one due to the reduced size of the new case.

The following are averages of three measurements each, taken from 6 inches in front of the case. This is favorable to the 550D, as it has two obstructions between the sound meter and the front fans, which reduce the noise. The V21’s front fan has a grill and filter, but no complete obstructions.

case idle db(A) load db(A)
Corsair 550D 42 43.8
Thermaltake Core V21 43.4 46.3

The idle noise increase isn’t noticeable. The load noise is lower with the 550D, but it’s not enough to matter if you’re doing something else that makes noise, like playing a video or game.

case max CPU (°C) max GPU (°C)
Corsair 550D 89.4 81.0
Thermaltake Core V21 71.4 74

The difference in temperature is quite large. For me, I’m going to take slightly more noise, by which I mean measurable but not really noticeable, as a trade-off for better thermal performance and cheaper, lighter computer.

I Will be Remeasuring Airflow

I discovered that I need to be more consistent about where I put my Testo airflow meter, so I created a jig from imitation Lego bricks. I’ve archived all my previous airflow measurements and so the bench will be empty until I upload new ones.

“Quiet” Computer Cases and Airflow

What is better? A computer case with sound dampening features, or a computer case without sound dampening features, with its fans turned down to match the sound produced by the quiet case?  Silent PC cases are damn heavy, and so what I’d like to believe is that the same benefits can be had by giving fans free access to fresh air while running them slower. I was fully prepared to trash my Corsair 550D, but it turns out that answering this question isn’t strait forward.

My case has two panels that can block sound and airflow from the front fans. Following you can see the case with all the front panels on, then only the inner panel, and finally the air filter. I leave the filter on for all my measurements for this article, but you can see the fans are behind it.

My prediction is that airflow will improve when the front panels are off. I can then reduce the speed of the fans to match the noise with the panels on. Airflow will improve enough to compensate for the slower fan speeds.

The computer was off for all measurements. I connected an external, silent power source to the fans. To measure sound levels I placed my generic meter in front of the case. I measured airflow using at Testo 405i with the sensor a few inches behind the fan. I tried to be consistent but I didn’t build a jig.

First, sound levels.

volts doors sound level db(A)
12 closed 41.5
12 open 44.0
8.8 open 41.5

The doors reduce sound levels significantly. What about airflow?

volts doors airflow m^3/h
12 closed 33.3
12 open 28.6
8.8 open 14.8

This is shocking and absurd.  How is it possible that adding obstructions increased airflow? Then it occurred to me that a spinning fan will move air, but it doesn’t mean that the air is moving from outside the computer case to inside. I entirely blocked the flow of air through the fans with cardboard to see how much airflow I would measure.

In spite of the cardboard, I found the airflow was 10 m^3/h. There is no way that the case is getting any more than 29 m^3/h of fresh air with the front panels on.

This experiment turned out to be more interesting than I expected, but I’m sad to not be able to conclude whether quiet computer cases, or at least the Corsair 550D, are worth it or not. On the one hand, my Corsair 550D does reduce the sound level from the front fans. On the other hand, to reach the same sound level when the fan was open to the outside, I had to reduce the airflow by half. But not the entire missing half is fresh air!

It’s entirely possible that merely half of the air moving behind the fan while the doors are closed is from outside the case. If so, then the sound dampening features are are useless. I suspect that the truth is somewhere in the middle: that the front doors do obstruct airflow, but not by half, in which case they have some limited benefit.