True. I don’t think efficiency wise AIO would be significantly better. I think how the fan speeds happen to be regulated on AIOs would actually make a difference. Using the water temp will have a more damped response than via the CPU silicon temp.
I have an idea — stick a temp sensor to the heat pipes of my air cooler. Use its reading to control the fan speed. Easy for Arduino to do this.
I have a NH-D14 in my old gaming rig I use as a Plex server. I use the Ultra Low Noise adapter with it and I can say you don’t hear a damn thing. The NH-D15 from what I have seen should just be an upgraded version of the D14. So I would assume it would work fine, with the low noise adapters.
BUT @Joro does make a valid point. I would maybe look at adjusting your fan curves. As this would be a FREE way of potentially solving the problem. If this doesn’t work then the D15 is always an option.
On fan tuning, and you alluded to it a little in your original post, but it’s the ramping up and down that is annoying, not fan noise in general.
So as well as tuning fans so they don’t spin up as sensitively, you can also go the other way at the same time, and have them spinning faster by default even if you don’t need them to.
So the default fan curve for me had my fan idling at 800rpm, which is very quiet, but for my particular room and case I worked out I don’t notice my fan noise until ~1600rpm or so anyway, so I set the min to that. Now my curve is something like <65ºC = 1600rpm, <75ºC ramps to 1800rpm, <95ºC ramps to max.
This has basically eliminated cpu fan noise for me, and I’m using the stock cooler than came with the 3700x.
Update: I have tuned the fan curve and I’m happy with the result
My observation is that when the silicon is hot, the radiator is usually only warm to the touch, so it means the bottleneck of heat dissipation is somewhere between the die and the heat sink. This means ramping up the fan would not help.
Also my original problem was when I scroll up/down a webpage quickly, the fan would ramp up and cause the noise. For this kind of transient load the only thing that helps is the heat capacity of the radiator, not fan speed.
Transient load usually wouldn’t push the CPU temp to more than 50ish degree.
The fan is quiet at <~1.2kRPM. So, I adjust the fan curve to an exponential one, to have smoother transition at lower speed, with 50ish degree below the audible fan speed.
I will come back and post a screenshot of the fan curve
This is the thing I severely dislike with “noisy/non-noisy” qualifications on the internet. The “acceptable” level of noise is subjective. I consider my noctua nf-s12a “loud” above 600rpm.
If only there was a like a phone app to measure noise (there is), and some household friendly way to calibrate it (between different phones in different cases). That way we could make actually applicable objective measurements without expensive test equipment or trying to remember or imagine how noisy roughing leaves or jet engines are.
Let me clarify. That’s how I worked out my settings. Other people can use the same method (not the same numbers!) to adjust their settings accordingly. Make sure the flat part of the curve sits below the noise tolerance threshold.
Yes it has to be subjective. It depends on each person’s setup, environment and tolerance level. Unless it’s fully passive with absolutely no noise…
You never ever ever want to have 0% fan speeds…
That jump from 0RPM to whatever RPM 40% (pfff it’s almost 1000RPM) is is gonna be noticeable every time. What you want is constant speed that is low enough that is inaudible and does not change. In my system that speed is constant 30% fanspeed. Also keep in mind 50C for a CPU is basically ice cold, so you can glatten the curve. You want the fans to start changing speed above 60C, because in basic every day browsing, videos, Youtube stuff your CPU should never reach 60C so your fans would never change their speed…