Every time I begin thinking to myself “maybe I’ll stick with my Batch 14 Desktop order instead of my Batch 15 Mainboard order, only so I get the ultra-SFF case and live with the PSU fan” … I’m reminded in sobering fashion that I have a Blackview MP60 sitting on my desk in-front of me. That I stopped using with the intention of taking it apart to try to alleviate the Mac Truck-like sounding fan that it has developed. So go back to knowing that I’ll be thinking and doing exactly what you thought and did, egalanos. Thanx for the reminder.
BTW, my guess is you should probably make a recording and open a ticket. My belief/assumption is there WILL be a correction of the PSU fan behavior by Framework, and “hope” they’ll offer a swap-out replacement program to their customers. I assume to be eligible for that fix in whatever form it takes, you’d want to have a ticket open.
I’ve solved this by flipping the power supply over 180 degrees to put the wires next to the main board. Then the duct interferes with two of the mounting screws for attaching the duct to the PS, so I’ve made it two pieces with a bayonet mounted duct. It makes sense not to do all that for production, but with a 3d printer and some patience, it’s a 50 cent mod perhaps.
So far I have this. It was just kind of tossed together today, not really ready for prime time yet. This is a 32mm exhaust, the 42mm was pretty massive, bigger than the vent on the case by quite a bit.
The plate has four keys to hold the duct on. The duct is printed in two pieces, the duct itself and a small ring with four keystones that I glue to the bottom of the duct. The extra hole in the duct was an attempt to get access to the screw on the plate before the attachment solution of making the duct removable and redirectable.
It needs bigger locking keystones, they are just barely big enough to print on an FDM printer right now, and are really pushing it. They don’t have to be particularly strong since the duct isn’t dealing with any forces, but it is very delicate right now.
You probably could massage the wiring under the duct carefully and put it in the stock position, but the framework case power supply grill is only 22mm diameter so it needs some modifications too. The grill is also beefier than it should be, to keep it from breaking I’m sure, but it is blocking a lot of airflow.
“This is a hardware circuit that turns the fan on when the internal temperature of the power supply exceeds a certain threshold. During low system load, the fan stays off, and once the threshold is passed, the fan gradually ramps to the target RPM.”
My fan is audibly instantly going from off to a high speed, not ramping as described.
“We spent several sessions inside of an acoustic chamber listening to and measuring different Delta fans to choose the one that sounded best while pushing both enough CFM and enough pressure to handle full load at high ambient temperature.”
In my opinion the fan has a pretty annoying tone, but that’s subjective.
What’s most annoying about this is that the fan cycles on and off in sleep mode, making it noticeable at night. I’ve seen the BIOS option to shut off the power supply while in standby, but having any USB device plugged in that draws more than a couple of watts appears to be keeping the power supply on in sleep anyways. Ex, RGB keyboard, external SSD, etc….
I could make the PSU fan silent letting the heat dissipate at the bottom of the housing.
I designed and printed a bottom frame with legs. If you have 3D printer, you can make it at home, otherwise try lifting the device with 4 bottle caps or similar, and see how much it helps you to reduce the fan activity.
I have attached the 3D model files. You can glue the frame and the legs together.
Hope it helps You at least a bit. I am running Ollama and FastFlowFM LLMs without the PSU fan becoming noisy on Windows. No experience with linux yet.
All the best!
I wonder if there could be a more « aesthetically integrated » version of this hack which take upon the front tile design to have four little cubes the same size and finish as the tiles, and aligned with them ?
I am batch 13, so I guess I will soon have an opportunity to find out.
Intriguing! I’m wondering if a similar result could be achieved by having the machine on its side or - gasp - upside down? Is there any reason not to do that?
Awesome idea! 15 min. after reading your post I had thrown together a make-shift frame from wood scraps. Just four 2 x 1cm sides in a rectangle that is smaller than the actual chassis footprint (was just eyeballing it).
It might need some legs though - currently it’s basically trapping warm air, but should in theory still be better than before. After running for some time I already had the PSU fan turn on, but it seems to turn on for a shorter duration than before. I bet after a cold boot it will also take longer until the first time the PSU fan kicks in. Might try to build a better one on the weekend. Waiting for a bit to see what a community based 3D printed solution might look like in a few weeks before ordering anything custom printed.
Full disclaimer: I do have the main fan constantly running at 15%, as that seems to have positively increased the time between each time the PSU fan turns on. Happy with any tiny improvement - every small hack adds to a better overall experience.
Very disappointed that there is still no official communication. Framework leadership again dropping the ball there
Here is the comment I added to when I canceled my Desktop order but kept my Mainboard only order instead:
I will be very unhappy with the PSU fan behavior of 0 to 100% rpm every few minutes and the audio spectrum that it produces as a result. Wish your supplier hadn’t screwed you and undermined your efforts for a quiet system, but they did. So I am going to miss out on the great SFF case as a result of that dang PSU you’re supplying ATM. Hope that changes!
Good luck Framework. I hope you can claw-back some financial concessions from the PSU manufacturer. But I imagine, they don’t care. And that sucks for you and for us.
Unfortunately for me, it does not really help with the problem. Possibly the fan spins up at slightly longer intervals, but it still spins up regularly.
Oh, that is nasty! I’d say that’s a dead bearing as fans should just make a blowing noise and never buzzing, but of course recordings can be misleading.
The CPU also releases heat, which goes up as a warm air flow before the CPU Fan kicks in, what barely happens with normal daily office use. So it might do things even worse, try it out and report back your experience. Why not ? Maybe we ourself become F1 Aerodynamic and entropy testers at the end . But sad to say that we have to take care of such things.
Madness but maybe we could drill many small hole(s) on the PSU side/a big whole and somehow flow air in/out from the housing, more than the actual solution. But that will be definitely ugly. Framework/supplier actually put an external laptop-like power supply into a closed house. They tried to release heat adding a built-in fan. That is loud and requires to spin on higher rpm to produce the required airflow due the small form factor. On a normal size PSU there is enough fan and outlet diameter for low rpm, but not here in this tiny brick. If someone has an industry thermometer to picture the situation when the PSU fan is idle and spinning on high rpm for comparison, we’d be smarter to take actions . I am thinking why not to remove the current PSU and replace it with an external one without fan, like a laptop…
I don’t notice the noise others have called out. I can hear the difference in the audio samples posted by a couple of members here, I just don’t notice it in my environment normally, a mostly quiet office where I work alone half the day.
The enlarged fan duct I modeled has been improved and simplified. It is two pieces and can be oriented in four directions depending on use case. I just don’t think it is the problem, it has done very little to change any behavior of the system for me. But perhaps those more sensitive to sound would notice something.
The root problem, as others have mentioned, is the inefficiency of the power supply overall. It just has too much idle draw, causing the case to run hot. To avoid a constantly on fan, the temperature curve seems to be set to allow the normal but elevated levels. Then the fan has to kick in when the temp quickly gets above that baseline during any load.
To compound that, the airflow inside the power supply is either very restricted, or just doesn’t flow past many heat carrying surfaces, in spite of the heat on the outer case body, the air exhausted out the back of the PS is cooler. It may be a sufficient PS to power the MAX+ 395 board and chip, but it was not engineered to do it silently. If you are sensitive to that, I’d say shop for a higher efficiency supply.
You could also try ducting the exhaust side a bit away from the desktop. Exhaust ducts of just a percentage of the fan diameter can often help muffle the blade noise.
I have attached the stl files for the fan duct if anyone else wants to mess with it.
When I tried this for the first time yesterday my machine was already running for a longer time and pre-heated. It did improve how long the PSU fan had to turn on each cycle, but it stilll needed to turn on.
As suspected it improved behavior after first boot of the day tremendously. Has not turned on for the first hour at all. Curious when it will reach critical temps, but it’s such a pleasure to work with a silent machine. At least for some time now - and still have not added legs to stand on which will most likely improve this further. Might even try to set CPU fan to auto instead of 15% base level.
Thanks again @AttilaBerencsi for this very welcome band-aid while Framework are hopefully figuring out a permanent solution.
. But sad to say that we have to take care of such things.
.