I got my 128GB delivered on 12/26, and have been slowly getting it set up. Its purpose is as a home server, so it’s powered on 24/7, but (so far) it’s been mostly idle. Even still, the PSU fan seems to be on more frequently than off. I still need to sit down and try to measure it, but instead of the pattern people have been mentioning of ~2 minutes on, ~10 minutes off, it feels to me like it’s almost inverted, like it’s mostly on, and takes short breaks. The computer has plenty of access to open air, is completely unobstructed on all sides, and I’m in Vermont with cold winters, average ambient room temperature of ~70F. I live in a small apartment, and I can hear the fan easily above ambient noise within an 8-10 foot radius, which basically means my entire apartment other than the bedroom.
The PSU fan turning on in the context of the constant on 20-30% APU fan, misses an operational domain, where the APU is being utilized at low 5-20% but the APU fan curve still stays low rpm.
This causes the PSU fan to spin-up and enter the dreaded 100% cycle. So it, the issue if you can call it an issue, is in the curve of the APU fan.
But above all running the APU fan is a stop gap mitigation. It is not a solution.
Was the same for me. 30% was the threshold where most days it does not turn on at all all day. Maybe on two out of 10 days, but hardly ever with my usual light work / watching videos.
Also tried 35% which was fine of course too, but the fan is much more audible then. Much less annoying as the PSU fan, but I prefer it more silent at 30%.
Framework sent me a replacement PSU, switching to the new PSU did not help.
Bit annoyed because I linked this thread in the support case and specifically asked them if they were confident the new PSU would help before they sent it. They didn’t really answer that, but were insisting this was something they needed to try to determine where the noise comes from…
Side note, but I’d be interested to know if there’s a different PSU that fits very well in the case. Not just form factor but cable lengths and so on. I found a couple which supposedly has customized cables to make them shorter (and one modular). But since this is fairly niche I’d really like to know that something works well before ordering myself. Anyone tried these or something similar?
I didn’t get lucky, and the replacement fan I got from Framework behaves exactly the same. I didn’t want to spend an hour disassembling the computer just to realise I had wasted my time (like everyone else here) so I plugged it from the outside. Then I realised I could actually measure that stupid fan not ramping up, so here goes nothing:
Removed the fan duct; put a white dot on the fan where the small grove is; recorded with a gopro at 240 fps (put shutter speed a ~1/1000, needs good lighting); then passed the video in a vibe coded opencv script. I’ll put that on github in case anyone wants to reproduce.
The problem is really that the PSU becomes a hot box even outside of the case as the fan doesn’t spin before having to go full speed. The only reason I can think of is to avoid accumulating dust there, given that we can’t clean it. Is that a thing?
For ref, Framework does say the fan gradually ramps up here:
During low system load, the fan stays off, and once the threshold is passed, the fan gradually ramps to the target RPM
At full speed it is fairly quiet outside of the case though, so maybe the acoustic tests they claim to have done didn’t happen with the full computer around the fan.
Thank you for the measurement, but to be fair, 2500 RPM is probably not the full speed of the fan. I don’t know exactly which fan it is from Delta’s catalogue, but usually 40mm fans can easily go beyond 6000 RPM at full speed.
That said, it would indeed be nice if the fan could run at lower speed than that, and even better if the PSU didn’t get so hot at low loads.
I disassembled the PSU and removed the fan.
In its place, I installed a Noctua NF-A8 80 mm fan.
Because it was 5 mm thicker, I decided to break off the rear part of the power supply housing—the section with the grille.
I cut the cables with scissors and, by twisting them together with my fingers, adapted the Noctua fan to the new connector. I completely removed the yellow wire.
After that, I more or less forced everything back in so it would fit somehow.
And surprisingly, it runs noticeably quieter. I can tell it’s working by the perceptible airflow at the exhaust. And now it’s very quiet.
Update on my Framework desktop, for the first few months the PSU fan never turned on (or I didn’t hear it), a couple weeks ago it turned on for the first time, and it’s unreasonably loud. Now the PSU fan is frequently turning on, even at idle. Since then I 3D printed the stand for under-case airflow, set my CPU fan to 30% minimum, and the PSU fan still decides to turn on at idle. Not sure why this happened, I am a little sad my PSU went from silent to problematic.
Well, and here I thought, I was the only one. During the first month - the honeymoon phase - I didn’t notice it as strongly, but the 0 to 100% PSU fan noise every 10-20 minutes or so from merely browsing is a bit annoying. I’m not young, being 40, but the whirr IS grating after a while. I probably just didn’t notice it as much, because I often also hear music, but during work I need my silence, and that’s when I notice that unpleasant problem. But after studying all the responses, it seems to be a problem in the PSU design. Shame. In essence, I love my small working horse, but between the unreliable drivers and the PSU noise… the regrets are mounting.
My PSU fan was making noise maybe every 20-30 min. It wasn’t the worst thing in the world but it was actually louder than the Noctua APU fan at full speed, and since I tend to leave my Desktop on even at night, it became kind of a problem.
I would say the problem was the PSU overheating, because the fan would start less often with the left panel open, and more often when the Desktop was pulling more power (during a CPU stress test for instance).
I ended up contacting the customer service about it. There was a lot of back-and-forth and troubleshooting. They asked me to send them videos of the Desktop making this noise, which I found reasonable, but then started asking more tests like with and without the USB modules (? what could this possibly have to do with the PSU fan ?). At some point they asked for a video of the PSU fan which I found a bit egregious (I would have had to take the whole computer apart and remove the PSU air duct I guess) but then they changed their mind and sent me a replacement part instead.
I installed it, it’s not completely straightforward since you do have to remove most of the internals to reach the PSU. So far (ie in the past hour), the PSU has been quiet. I ran a stress test, which previously would end up triggering the PSU fan (I guess because of the additional power draw) and I didn’t hear it. I guess I’ll see later if they actually fixed the issue or not.
EDIT: still noisy. It’s a higher pitch now, and definitely less audible than before, but still there from time to time.
As for the photos, I don’t have any better ones. I was too annoyed that I bought a fan that was too thick. I should have bought the NF-A4x10 instead of the NF-A4x20, so I had to improvise.
After that, it’s just the PSU itself. Disassembly is quick. In the guide above, steps 36–37 show the screws that need to be removed. The screwdriver included with the computer is sufficient for this.
The actual fan replacement and connecting it to power was a bit of a makeshift solution in my case. However, this video shows that it is also quite easy to do:
Hi, I received my Framework Desktop a few weeks ago. I am posting here to voice my feedback, similar to many others, that the PSU fan on the Framework Desktop is annoying and I have sadly relegated it to under my desk where the fan is not as noticeable.
The reviews I saw made it sound like this would be a quiet PC with the large Noctua fan and didn’t mention the PSU fan. I have a quiet room and the higher pitch PSU fan turns on every 5-10 minutes for a minute or so, even with the computer is at idle. The intermittent on and off is very annoying; it would be much better to me if there was an option for it to stay on constantly at a lower DB. The noise is especially annoying when the computer is on your desk a few feet away from you, which I was hoping to do because of its small size!
I see a lot of posts here with work around hacks but I don’t have a 3D printer and know I’d probably waste hours/days tweaking/obsessing trying to find a quiet solution. You can call me the “general audience” who just wants things to work and isn’t much of a hacker. There are many Framework customers who have fun hacking their computer, but I bought Framework more for the customer support and modularity… I like that I have the option to switch out to an upgraded mainboard in a few years which is not possible with other Mini PC vendors.
Instead of returning it I will keep it under my desk where it is not as noticeable, because the great performance and much lower TDP make it still worth it, but I see this as a major flaw in this product. Framework should treat this as a major issue!
I hope Framework can come out with an official statement and come out with at least some temporary fixes.
I hope they are working on finding a better vendor for the PSU that is silent or with much better fan and control settings. And I hope they make this available to current customers when they source it.
Good choice to keep it! A desktop with that much performance in that form factor is incredible. Love it too.
My guess is that one day they’ll just announce new PSU. Not sure how they’ll handle replacements, but so looking forward to that day.
@janne_anderson many opened a support ticket. Would recommend you do the same. You can tell them that it’s just to let them know and that you’ll wait for a solution. So that they can gauge how many people are affected. Just good in general to document it.
The actual team chimed in (not just 1st level support) and they are analyzing and trying to come up with a solution. It’s a bit more involved as they have hoped and it takes some time. No actual timeline or anything concrete is known at this time.
but not a good long-term solution