Performance Testing

Please refer to the dedicated test setup page here as it applies to this review as well and I did not want to go over it separately to save on review space.


As per usual, I hooked up the fan to my trusted Aquaero, set it to PWM mode, and started mapping fan RPM as a function of the PWM duty cycle. It was immediately apparent that something was off, however. The Scythe Kaze Flex 120 RGB PWM fan behaved quite erratically to where even going from 100% PWM to 99% PWM resulted in a massive RPM drop. This continued until ~95% PWM duty cycle, following which things settled down and changed slowly. But that is still going from an average of 1135 RPM to 485 RPM in the first 5% (for three test samples). I saw something similar with the be quiet! Silent Wings 3 PWM fans, and that ended up being an issue wherein be quiet! opted not to stick to the current Intel PWM spec to increase compatibility with motherboard fan control that—according to be quiet!—was still using an older spec with a different pull-up current. Indeed, I tried out two different motherboards for fan control and had the same issue, so it is not my Aquaero alone. For what it is worth, voltage control did seem to be working as expected, but voltage control on a PWM motor is not recommended.


This began a long back and forth with Scythe, including a visit to their facilities in Taiwan where they showed me how they test their fans using some motherboards and a power supply. The very fans I had tested different there, but they did not have the same controller or motherboards I had. They then sent over two fans with an updated motor based on our discussions and previous findings, and as seen above, these worked way better. Scythe says they do not plan to update all their fans to this spec given they maintain this issue is mostly with older motherboards, but given I also saw the same issue with an Intel Z270 chipset motherboard from a popular motherboard maker, I beg to disagree. Given these new samples are not going to be indicative of the retail fans, I used voltage control with the older fans for the rest of the tests.


Let's remind ourselves that the fan is rated for 300 to 1200 (+/- 10%) RPM, presumably in PWM mode. We saw above how that went, and the fan in voltage mode went from an average of 1135 RPM at 12 V to 336 RPM at 2.75 V before shutting down. All three original test samples restarted at 3–3.1 V, and the RPM range of operation was close enough to the rated specs to be permissible, especially given the different mode of control! Sample-to-sample variation was minimal as well, which is good to see.

Context is needed to talk more about the fan's performance and noise, so I have below comparison charts for some fans tested so far at set RPM values (or as near as they can get to those).


I have included fans in charts where the rated RPM is within 50 RPM of the chart cutoff point, which means that some fans are in specific charts only if their rated speed is over 50 RPM off from a threshold value (Corsair SP120 RGB, for example), or they simply do not slow down enough (NB-eLoop B12-4, for example). Similarly, the specific RPM values chosen reflect usage scenarios most popular with watercooling even though some fans (the Noctua NF-A12x25, for instance) go higher. The charts are to be considered for comparison within this result set only and are not to be compared with results from another test elsewhere owing to different testing conditions.

This is likely the last time you will see the Scythe Kaze Flex and a few other fans in the comparison list, as this fan in particular only manages to make it onto the 2/4 tested RPM ranges. Even so, it does not tell a good story for the fan when tested in a relatively high airflow restriction use case, such as the radiator I use for fan reviews. The Kaze Flex 120 RGB is just not good enough, with other fans being as quiet or even quieter and yet performing better at the same time. Even the recently tested Alphacool Eiszyklon Aurora LT RGB, which did not impress then, looks better here by comparison.