Probably the ultimate low noise fan cooling for Framework Desktop case

Framework Desktop
1

The post Noctua fan doesn’t fit gave me a little bit of inspiration. If Flar was able to install a 140 mm Noctua NF-A14x25r G2 PWM, why should I not try something similar?

I got a Noctua NF-A15 PWM years ago for an older desktop and somehow I never used it.

I knew it has the same mounting holes, so I just had to try it. It fit perfectly with the fan duct on top, although obviously that will not have the same effect as when installed on a standard 120 mm fan.

image
image1920×2172 672 KB

The other two fans are: 80 mm in front for intake and 60 mm at the back for outtake.

Using info found on Noctua website I got this summary:

NF-A12x25 HS 2400 RPM, 69.25 CFM, 28.8 dbA (note 1)

NF-A14x25r G2 1500 RPM, 91.58 CFM, 24.8 dbA

NF-A15 1200 RPM, 67.98 CFM, 19.2 dbA

1 - This is the fan that Noctua customized specifically for Framework, slightly more powerful than the stock one that you can buy anywhere else.

NF-A15 does not have the highest air flow, but is the most silent of all. You cannot find anywhere a quieter fan with a reasonable air flow.

The APU sensor shows 39ºC for an ambient of 28ºC on balanced power with some web browsing, watching a music clip on Youtube and a video processing in the background, with absolutely no noise from the fans although they all rotate at 60% duty.

I think this might work fine for people who don’t like or don’t want to hear much fan noise from their Framework Desktop case and still have a reasonably good cooling.

2

I don’t even have a Noctua, I only have cheap fans, but I played around with a lot of configurations in an open bench. If you want to saturate the heatsink fins so that the heatpipes are the bottleneck, the whoosh of air through the fins is the loudest thing you’ll hear by far.

Larger diameter + lower rpm is trading pressure for flow to simplify things. A different way to get higher flow + lower rpm through a restrictive heatsink is a thicker fan.

My apartment’s HVAC and such are quite loud so I run 92mm on a shroud to get more even airflow through the fins (almost no hub shadow), and have the option of cranking up the speed to get the last bit of cooling out. I bought a 3900rpm fan but going past 3000 doesn’t help, oops, now I’m soldering on a resistor to slow it down relative to my other fans on the same PWM controller. At 2000rpm the whine isn’t perceptible in my room but in a quieter space it might be a little annoying.

3

To each his own, all of us have different requirements to make us happy.

I like the highest performance I can get in a silent environment so I’m trying everything that gets me that. It would be nice to have a fanless machine with the capabilities of the FD, but we are not there yet. Maybe in a few years!

4

What I’m saying is, how are you going to get airflow through the heatsink with that giant fan? Big fans are for unrestricted flow.

Given you already have a fan hanging off the back, one idea is stack a second Noctua 120mm on top and run both of them slower. More pressure to push through the heatsink at lower rpm. Actually, using 2 fans, you can pick one that spins the opposite direction at the same speed to cancel out the angular momentum, that would be extremely efficient and let you slow them even more. E.g. Scythe Grand Tornado 120mm 2000rpm. Then because the flow is already very straight, you can put a big spacer between the fan and heatsink to let the air expand into the hub shadow, reducing the velocity through the edges of the heatsink, which further reduces noise and improves cooling.

Rated noise is at a certain speed in open air, engineering low noise is more than picking a fan.

5

Maybe I should have added to the last post that I also like to keep things simple.

For now, this setup produces an APU sensor temperature hovering between 35ºC on idle and 45ºC for typical home-office usage, only going above that during short-time resource intensive tasks.

Over-complicating to squeeze the last drop of performance is not my thing.

But I surely do appreciate your input!

6

I may be wrong, but your claim that “Rated noise is at a certain speed in open air” does not hold water. Different manufacturers use different testing procedures to benchmark their fans and they don’t seem to publish how they do it. Which means that we have to take their numbers with a grain of salt.

Also, please see this, including the pdf in Section A. It seems that really professional fan testing is not done in open air, but in specialized chambers and/or wind tunnels.

Correction: It seems that you were right about the open air testing done by manufacturers. That produces the lowest numbers for dbA and that’s what they want.

7

My only spare 120mm is a crappy 3000rpm Delta, I made a 3d printed stator with a hub fairing (seems successful, no extra noise and the hub wake is smaller, air flings outward at a smaller angle), and tested it against my 92mm. The fan has some issues with backflow, but even at full speed I could feel a pretty big difference in air shooting out the sides of the heatsink in favor of the 92mm. I’m a little tempted to buy a cheap anemometer…

I took my own advice and ordered 2 92mmx15mm fans, 11 blades 2800rpm and 9 blades 2700rpm. In theory, a contra rotating pair like this should produce ~double the pressure at a given flow rate and speed, so I should be able to run it at something like 20-25% lower speed to get the same flow. There would probably be a net 1-2dB reduction in noise at the fans, but it’d be lower pitched, and the straight flow would perhaps reduce the noise generated in the fins.

My hope is that my rig can reach maximum performance +1C with 2 120mm intakes at 1600rpm and the pair of 92x15 fans at around 2000rpm. I made it harder on myself than if I used a much taller shroud and a 120x50 contrarotating fan, oh well.

8

You should be very lucky to have those two fans absolutely identical, rotating at the exact same RPM and giving you equally the same flow so their contra-rotating produces the straight flow effect that you want. I think I read somewhere a while ago that no two fans from the same line of the same manufacturer are perfectly identical. There are ALWAYS minor differences which, actually, may be valid for any product on the market!

Now let me ask you something: did you ever give some thought that all the resources, money and time you spend on this endeavor might outweigh (by a lot!) the benefits that you get? Or you simply don’t care?

9

I’m also thinking of trying something new, but only because the FD will not be my main machine soon, that role going to a much more powerful (and large!) AMD desktop.

So, I’m planning to go to the other extreme of quiet: I will install a Phanteks T30-120, which is one of the most powerful fans on the market, third to a couple of Thermalright fans that I cannot get in my location anyway. And at the same time one of the loudest!

It is a 30 mm height fan, instead of 25, so the top of the fan will touch or be very close to the side panel, eliminating the need for the Noctua designed fan duct.

10

You don’t necessarily want perfectly synchronized fans for noise reasons, the momentum cancellation is almost perfect with a small difference in speed.

These fans are all cheap, it’s just a hobby. It’s for the love of engineering. I added pleats to my side intake filter today to let it breathe a little better.

11

Good for you! I’m also an engineer, but I learned early that when you try to minimize something, anything, something else goes the other way. For me too much effort becomes counter-productive and it’s not worth it. For some, like you, the fun of it overcomes that and keeps it an interesting world!

12

I also added a 60mm USB external fan on the top and a smaller one on the bottom for the PSU, works great so far Extra cooling on FW Desktop

13

On my FD the fan on the PSU very seldom comes on. It looks like a good cooling in the case eliminates the need for that little guy to start squeeking.

14

Have the contrarotating 92x15 fans now, not totally happy but this is what I’m going to run as I have case constraints (side intake centered for ATX not ITX, and side panel is already restricting the fan). Because I stacked them on top of each other (fan supports facing each other, so already maximum rotor distance), the blade interaction noise is quite significant (both a loud whoosh and a resonant hum when the speed gets higher).

If I were able to give them 10mm more spacing, they become very quiet, but as I don’t have space I glued them together with epoxy putty (forming a smooth center tunnel). I can hear a bit of difference even with a silicone gasket in between them.

The pressure increase is as good as expected, I can feel a very strong air current coming out of the heatsink comparable to the old fan running at a much higher speed. Overall noise at the same temperature is definitely down. At max power I have the fans running about 2100rpm and I only hear a soft white noise.

I think if you have a lot of space to spare, the perfect setup would be something like fan → 20mm spacer → counter fan → 30mm shroud → heatsink. You could get maximum cooling at very low rpm. Or perhaps fan spacer fan tube 90 degree turn heatsink for a more normal computer case shape.

15

Since I decided to trade noise for CFM and I got the Phanteks T30-120, I can say I’m impressed:

The T30 has a 30 mm height instead of regular 25 mm for standard 120 mm fans, so the blades are wider and push more air. That 5 mm extra height also completely replaces the Noctua designed fan duct that is supposed to guide the air better on the radiator fins.

It is a monster of 101 CFM with a 40 dBA noise to match at rated 3000 RPM max.

In reality I recorded a max speed of just above 3600 RPM. The fan curve in BIOS is set with a duty of 40% so it runs at around 1360 RPM with FD at idle with a slight woosh. I also set the Noctua front and back fans at 90% since they make barely any discernible noise.

With this setup I get an 8ºC delta over room temperature. It is quite hot in my location now, so the FD APU sensor sits around 39-40ºC on idle and it goes to 75-80ºC on short duration demanding tasks. During my typical work on the machine and up to 60% RPM of the T30 the noise is bearable, after that well, that’s life!

1 Like
16

Mounted right against the fins, a thicker fan spinning slowly should work well. The only downside is the flow pattern, more pressure being generated at the outer edges of the fan, dead spot in the center.

My dual 92mmx15mm fans are able to saturate the heatsink despite a bad looking CFM rating. I can get identical temps to the 4000rpm 92x25 or the 3000rpm 120x25. I reapplied thermal paste and now am at ambient +53C 140W load :slight_smile: Fans on whisper quiet would raise it by around 5 degrees.

17

I get about the same results with this thicker fan: 50-53ºC delta over ambient at 140 W load.

Still waiting for the pair of 30 mm fans to add for outtake at the back. I tried two brands bought from Amazon, one was too noisy and the other one was silent but weak. Last I ordered Sunon, supposed to have 4.6 CFM at 20 dB. Adding 9 more CFM to the outtake can only help and they will be the last improvement I’m doing on this machine.

18

hmmm im new, going to order :money_mouth_face:

but got a little confused reading this topic about fans. :open_mouth:

the order page has the 120mm fan option, so i assumed there is only 1 fan. then i saw the phoronix page which shows another 80mm fan at the side. but it is inside. then i saw the photo above, there are 2 fans :open_mouth: . so i got curious and installed freecad to look at the chasis, and had a peep at the schema. and there inside is 3 fan headers :thinking:

assuming the block diagram is right, the +12VB_SYS line max supply = 4.6A, feeding PCie (=2A) and 3 fan headers. so there is about 2A headroom for the fans.

the CM/noctua 120mm consumes about 0.1A, does about 60+cfm/20+Pa. this power optimized under 2w. on mouser, there are the 5-7w range fans with “better air-watts” (which still fits the supply rail envelop) going at >100cfm.

are the side fan/fans really required ?

*edit : the 80mm fans are in the range of 40cfm. the crazy >5000rpm do more cfm / noisier. so why not have 1 good 120mm do all the work for less watts?

19

That’s exactly why I replaced the original Noctua 120 mm with the Phanteks T30-120, it’s much more powerful, albeit louder at high RPM.

I went through an intermediary stage of Noctua 140 mm fan, which is also a good option if low noise is important, but you will get slightly higher temperatures than the Phanteks.

You should not worry about the amps of the fan headers, they are rated for 25 W each and the typical fan draws 4-5 W.

The front and back fans (you call them side fans) are not required, but they improve airflow a bit.

To me, the main purpose of the front fan is to cool down the third SSD that I added to the PCIe slot, using a PCIe to NVME extension. The SSD sits now right in front of that fan.

20

Depend on the design of fan blades. There are “airflow” and “static pressure” fans, choose the latter.