Raspberry Pi 4 – 40mm Fan Comparison
The current start performer of the active cooling tests is the John Sinclair 3D printed case, see the full review. This case features a top-mounted 40mm Fan which does a great job of keeping the Pi cool, even under full stress conditions. The supplied 40mm fan is fairly quiet, but not as quiet as the fan used on the Pimoroni Fan Shim. I wanted to see if I could improve this already great actively cooled case by testing a handful of different fans.
Selection
As supplied from eBay the kit fan is a CUI CFM-4010V-070-273 (datasheet) since I’ve had positive experiences with the Sunon MagLev series (catalogue) of fans in the past, I wanted to compare a similar spec fan the 7000rpm MF40100V1 against the CUI fan. I also wanted to look at the slower and what should be quieter MF40100V2 version. I had to include Noctua, as one of the best quiet high-performance PC fan brands available.
Spending more on a Fan than the cost of the entire 3D printed case and stock fan may not make a whole lot of sense (John Sinclair has confirmed he can, and has, supplied cases with this premium fan installed). Finally, I wanted to include what has seemed to be the quietest fan in my other testing. The Pimoroni Fan Shim, which uses an AVC DATA0307R5H Fan (catalogue). This is smaller, at just 30mm than other fans and is rated at 3.3V volts compared to the 5V of the other fans, so we’ll try and compare how well it performs installed in the JS 3D printed case.
| Make/Model | Current (mA) | Speed (rpm) | Air Flow (CFM) | Pressure (inch H²O) | Noise (dBA) | Price (£) |
|---|---|---|---|---|---|---|
| CUI CFM-4010V-070-273 | 157 | 7000 | 8.0 | 0.19 | 27.3 | 3.51 |
| Noctua (3pin) NF-A4x10 5V | 50 | 4500 | 4.8 | 0.07 | 17.9 | 11.49 |
| Sunon (MagLev) MF40100V1 | 136 | 7000 | 8.0 | 0.19 | 27.3 | 3.12 |
| Sunon (MagLev) MF40100V2 | 94 | 5800 | 7.0 | 0.12 | 20.6 | 2.58 |
| AVC DATA0307R5H | 60 | 4200 | 1.85 | 0.034 | 18.6 | 9.60 |
Testing
These 5V fans should be able to run at 3.3V too, this lower voltage will lead to lower RPM and reduced noise. So as well as testing 5 different fan models, where possible I’m testing them in 5V and 3.3V powered configurations.
I’ll be performing the same staged stress testing I’ve used previously, measuring the thermal performance at 25%, 50%, 75% and 100% load using Stressberry. Each test running for 1Hr. The test will factor in changes in ambient temperature to provide a temperature measurement above ambient that can be compared across all scenarios without needing to worry about environmental changes.
Along with the thermal testing, it’s important for this test to compare noise levels too. This is the first time I’m attempting this, but the method is simple. For each Fan and Voltage level tested I’ll use the same smartphone positioned in the same location 1m away from the system under test and take measurements of both noise level and hopefully identify the approximate frequency of the noise.
Heatsink
Previous testing has shown that with active cooling a heatsink provides a benefit, due to providing an increased surface area for the airflow to interact with. All tests with 40mm fans used the excellent HeatSink supplied with the JS 3D Case.
When testing with the Pimoroni Fan Shim the height of this heatsink was an issue, especially as I needed to access the 5V GPIO header to power the ambient temperature sensor. As a result testing with the Fan Shim is without a heatsink, but the fan is mounted a lot closer to the CPU and board.
Fan Installation Notes
The stock CUI fan is just pressed in to place on the top half of the 3D printed case. The Sunon MagLev fans fitted easily, with about the same amount of pressure to fit. The Noctua fan’s housing appears to be slightly larger and was a lot harder to get it pushed into place. I was concerned that I might break the case in the process. I’d suggest if you’re thinking of using a Noctua, ask John to supply it as I suspect he as a slightly different 3D printing file he’d use to ensure it fitted correctly.
Temperature Results
Temperatures in all tests with active cooling kept the Raspberry Pi 4 sufficiently cool at all time to avoid hitting thermal throttling triggers.
Fans @ 5 Volts
1 Core 2 Cores 3 Cores 4 Cores
On paper, the stock CUI and Sunon V1 fans should perform the same. However, the Sunon MagLev V1 fan has a slight edge over the stock CUI fan. Performing the best in all the test at 5 Volts. The slightly slower V2 version of the fan performing similarly to the V1 version.
The Noctua is the most expensive and slowest RPM fan in this test, this is reflected in a higher temperature in all tests, but still with a significant margin on the thermal throttle limits.
Fans @ 3.3 Volts
1 Core 2 Cores 3 Cores 4 Cores
Powering the 5V fans with just 3.3V leads to lower than spec RPM. This results in the Stock CUI fan performing at the same level as the much slower Noctua fan and the AVC fan of the Pimoroni Fan Shim.
Optimal cooling at this voltage is achieved with the Sunon V1 and V2 fans, which perform very similarly, from a thermal perspective.
Noise Testing Results
The noise produced by an actively cooled Pi enclosure can be an issue, smaller fans tend to be more problematic. The 3D printed enclosures support of a 40mm Fan should certainly help keep things quieter than a 30mm equivalent. Along with the properties of the Fan, the design of the case can significantly impact the amount of noise produced.
Initially, I planned to test with the microphone @ 1m from the Raspberry Pi. However, for the quieter fans, it was near impossible to measure a difference above the ambient sound level. So reduced the distance to 10cm, with the Microphone facing the end of the case where the SD slot is located, so that there wouldn’t be airflow directly towards the microphone.
Tests were performed using DecibelX for 1 minute, using ‘A’ Frequency Weighting, dBA. The baseline/background sound level was measured ahead of the test at 25.2 dBA.
Note: Reducing the house background noise level and keeping everyone quiet was probably the most challenging part of trying to perform this test. I eventually resigned myself to needing to perform this late at night when everyone was asleep.
| Make / Model | Voltage | Average Sound Level (dBA @10cm) |
|---|---|---|
| CUI – CFM-4010V-070-273 | 5V | 49.1 |
| CUI – CFM-4010V-070-273 | 3.3V | 37.9 |
| Noctua – NF-A4x10 | 5V | 38.2 |
| Noctua – NF-A4x10 | 3.3V | 27.7 |
| Sunon – MF40100V1 | 5V | 51.7 |
| Sunon – MF40100V1 | 3.3V | 40.4 |
| Sunon – MF40100V2 | 5V | 47 |
| Sunon – MF40100V2 | 3.3V | 35.7 |
| Pimoroni Fan Shim | 3.3V | 39.9 |
From a noise level perspective, the Noctua is a stand out performer, at 3.3V it’s near impossible to hear. Yet, as shown still provides sufficient cooling to avoid any thermal throttling. If noise is important to you, then the Noctua 40mm fan running at 3.3V is as close to a passive cooling solution as you can get.
The stock CUI fan lands in between the Sunon V1 and V2 MagLev fans from a noise level perspective, though I have noticed my CUI fan has generated a subtle variation in its noise output, indicating some change in the bearing, no maybe a new one would a better result. Given the slight improvement in thermal performance of the MagLev fans and slightly lower costs, the Sunon V2 version looks to be marginally better than the current stock fan. Maybe with this feedback, John might change what he is supplying.
Conclusion
Based on your primary requirement fan selection can be presented as:
- Thermal performance: Sunon MF40100V1 @5V
- Minimal noise (no cost limits): Noctua NF-A4x10 @3.3V
- Balanced option (thermal/noise/cost): MF40100V2 at either 5V or 3.3V
Part Suppliers
Searching for sources of the Sunon MagLev fans was a bit harder than I expected. At the time of purchasing, I wasn’t just after the 40mm versions but was also investigating a number of 30mm options. The usual UK stockists didn’t have what I wanted either due: only offering old discontinued version, stock on backorder, on just not listed. I found that I could source them from Germany from https://reichelt.de whom I’d never used before but offer a huge range of components.
- Raspberry Pi 4 Model B (Amazon)
- CUI CFM-4010V-070-273 (mouser) (digi-key)
- Noctua NF-A4x10 5V (Amazon) (reichelt.de)
- Sunon MF40100V1-1000U-A99 (reichelt.de) (digi-key)
- Sunon MF40100V2-1000U-A99 (reichelt.de) (digi-key) (mouser) (Amazon)
- Pimoroni Fan Shim (Amazon) (pimoroni)
You’re pushing the air down on the CPU, correct?
Yes, the fans are orientated so that they are blowing down on to the CPU
i bought the case and a Noctua NF-A4x10 5V (not the pwm one, not the FLX one)
how can i run the noctua on 3.3 V ? it is too loud atm on 5V (we are sleeping in the room where the Raspberry Pi is)
I imagine right now you probably have the fan connected between pins 4 (5v) and 6 (Ground) as they are next to each other so likely fit the standard connector on the Noctua fan. In order to power it from 3.3V, you’re going to need to connect the fan between pins 1 or 17 (3.3V) and ground on any one of the ground pins (6, 9, 14, 25, 30, 34 or 39). To do this I cut off the original connector and using created 2 single wires to Dupont connectors like (https://amzn.to/2sQyr4M). If you don’t have a need for a full set of connectors, you likely have some jumper wires with female Dupont connectors which you can use and solder these to the wires you cut on the fan.
Hope this helps.
How does this work with the max current of 50mA on the Pie at 3.3V? The Noctua has a max current of 50mA at 5V.
Hi Lars, The Noctua fan uses 53mA @ 5V and 33mA @ 3.3V (just measured this). So this is well below the recommended maximum of 50mA from the 3.3v rail. Had the fan been able to run at the same speed at 3.3V as 5V then yes it would have needed to draw more current. However, that’s not the case as this is just a simple fan, so at lower voltages, they just run slower and therefore quieter.
Just wanted to say thanks again for all the testing Martin!
Have had a few people ask me about the power available from the 3.3V rail & if running a fan from it exceeds the recommended current draw too!
Did a little digging and its worth mentioning that the NCP1117 regulator used on the Pi3/4 is capable of 800mA at 3.3V (admittedly shared between other devices on the board too), so even if the fan was to draw 50mA from the rail there shouldn’t be any issue. As far as I can tell, the specifications from raspberry just haven’t been updated since the Pi 1/2 which used a lower power regulator!
Cheers John for the extra details.
Hi Martin,
thank you so much for your work on this! It helped me a lot finding the right fan. Since my pi is located in the living room and running 24/7, both picking a reliable fan and eliminating noise as much as as possible was very important for me.
After reading you comparison, I gave the Noctua and the Sunon MF40100V2 a try. I ended up using the Noctua at 3.3V since the Sunon was still clearly audible at 3.3V when the room was quiet. On top of that, I didn’t like that the Sunon produced a bit too much vibration for me.
The Noctua, on the other hand, is just silent and still provides sufficient cooling. I’m using it in an open case (just an acrylic plate on top and bottom) and the difference in cooling between the two was even a bit less than with your setup using the JS3D case.
So again, thanks a lot for the great comparison!
Hardy
Thanks for the comment (and of course, for the article too) – incidentally I am nowdealing with and looking at exactly the same – bottom and top plates and either the Sunnon or the Noctua on Pi4. But I guess I will go ahead with the Sunnon first, my purchases have grown as they already are. I actually have a case with stoc han coming too, so we will see. Anyway, even if I am eventually ending with the Noctua, those cheapies will find use elsewhere.
The air flow is wrong for the Noctua NF-A4x10 5V in your first chart. Unfortunately it’s 8.2 m^3 / hour and not 8.2 CFM.
Apparently it’s only 4.8 CFM.
(didn’t check the numbers for the other fans)
Sources:
https://noctua.at/en/products/fan?size=2645&voltage=5
https://www.convertunits.com/from/cubic+m/hr/to/cfm
Thanks for spotting that. I’ve corrected the table for the Noctua fan and confirmed all other vendors data is correct.
Is there a case for a Raspberry Pi 4 that will fit the Noctua NF-A4x10 5V fan? My case is from Canakit and it came with a very loud fan. Unfortunately I ordered the Noctua but it doesn’t fit the case.
Hey Mike,
Most cases are only designed for 30mm fans. 40mm fans look to be more common in custom 3D designs like the JS3D cases I used: https://www.martinrowan.co.uk/2019/09/john-sinclair-3d-printed-raspberry-pi-4-case/
Honestly, unless you have good reason you’d be better returning the Noctua fan and instead, getting a really good passively cooled case. Like the Argon Neo or FLIRC. These cases can even cope with overclocking and you’ll never have issues with noise.
Thanks Martin