I have recently booted up a new CRS320-8P-8B-4S+ which currently is sitting in a rather cold room (below 15C) with 1 PoE-cam connected (802.3at) and it has OK fan speed on all internal fans (750-1100rpm) besides the psu1-fan which is running stable at 5400-5500rpm, even though it is at 20C.
0 poe-out-consumption 2.1 W
1 cpu-temperature 38 C
2 fan1-speed 720 RPM
3 fan2-speed 1020 RPM
4 fan3-speed 750 RPM
5 board-temperature1 22 C
6 board-temperature2 12 C
7 poe-out1-voltage 54.7 V
8 psu1-state ok
9 psu2-state not-present
10 psu1-power 17 W
11 psu1-temperature 20 C
12 psu1-fan-speed 5480 RPM
13 psu1-input-voltage 241 V
There are no SFP-modules inserted into the unit. The switch is running ROS 7.18.2.
Is this really normal? I cannot see why it would need to be running that fast when there is no need for it, is it possible to affect this? I’ve read on this forum, and other places, regarding high-RPM fans but I haven’t seen anything related to the PSU-fans specifically.
This is normal behavior. For most PSU fans ~5k RPM is their minimal operating speed. These are low-resistance fans designed to spin fast with low noise and low power consumption
I'd like to chime in - a way to better control the psu1-fan-speed would be much appreciated. In near idle settings, the chassis fans can be throttled down quite well, but the psu1-fan sits at > 3000 rpm without a real need to be that loud.
[admin@rack-switch] > /system/health/print
Columns: NAME, VALUE, TYPE
# NAME VALUE TYPE
0 poe-out-consumption 0 W
1 cpu-temperature 56 C
2 fan-state ok
3 fan1-speed 210 RPM
4 fan2-speed 210 RPM
5 fan3-speed 210 RPM
6 board-temperature1 46 C
7 board-temperature2 35 C
8 poe-out1-voltage 55 V
9 psu1-state ok
10 psu2-state not-present
11 psu1-power 19.6 W
12 psu1-temperature 37 C
13 psu1-fan-speed 3140 RPM
14 psu1-input-voltage 231 V
PSU fans are not used only to cool that particular PSU, they provide air flow for cooling other components, such as CPU/ASIC, PoE controlers and SFP cages/modules.
In output you posted it can be seen that PSU temperature is 56°C ... which might cause higher fan speed. And there were cases where CPU heat sink did not have proper connection with CPU (thermal paste not applied properly). Or there's some SFP module installed which draws lots of power and generates lots of heat (e.g. MT's own RJ45 SFP+ module) ... which causes fans to revv up.
Indeed. I've had some trouble spelling all those acronyms ... I believe I only missed 1 out of 13 ... so success rate higher than 90% (which is my internal target threshold ).
the psu(s) is/are physically separated by a metal baffle (or partition) from the area where the CPU, other chips and SFP cages are.
Cannot say for sure how fans are enumerated, but presumably fan1 and fan2 are the ones at the back and fan3 is the one on the side (likely specifically added to cool the SFP cages when one or more little stoves are inserted into them).
Anyway I believe that PSU fan speed is "outside" of what RouterOS can (or could) regulate.
If you are into experimenting, you could try checking which kind of fan is in the psu (three wires, four wires?) and which voltage it uses and see what happens with a "slowing down" resistor like (example only): https://www.aliexpress.com/item/1005004974264597.html
But - generally speaking - that psu has a lot of "oomph" (600W) in a really tiny space and 40 mm fan really need to go fast to provide some decent airflow.
If you compare it with a common ATX power supply in the same range of power, it is double or triple the volume, and it has usually a 80 mm or 120 mm fan.
Maybe (but it is a loooong shot) a mod could be (since you only have one PSU) to use the space for the redundant one to add a fan (pushing air in, since I believe that the PSU fan is blowing air out).
In my original screenshot, PoE power is 0 and no SFP+ adapters connected.
My gut feeling is also that the PSU fan is not controllable (at least in the same quality as the chassis fans) by the OS. The PSU possibly just does its thing and signals the OS its fan speed.
Adding a resistor (or similar hardware measures) would slow the PSU-fan globally - my concern was just during idle periods where the larger airflow is not needed.
