HAP ax3 CPU temperature is 50-53 degrees Celsius, is this normal?
The device is under 1-3% load.
The room where the router is is 22-23 degrees Celsius.
> /system/health print
Columns: NAME, VALUE, TYPE
# NAME VALUE TYPE
0 cpu-temperature 50 C
50° is fine for a passive cooled device, it obviously also depend on the temperature in the room it is placed, 22-23° is a bit high.
I would have said 46-48° with 18-20°, so the difference around 28-30° is the same.
The real question is how high the temperature raises when the CPU is fully or nearly fully loaded.
It is in these cases that shouldn't be excessive, anything up to 70° or so should be fine, it is when it stays over 75° that I would start to think of cooling problems.
Stock thermal pads and thermal paste aren’t the highest quality. Aluminum radiator is small, much smaller than full enclosure one on RB4011. Anyway I replaced thermal pads abd paste on AX3 and now temperature sensor showed me good results, dropped from +59C to +50C under the load with 35 wireless clients and ~30 firewall rules
It's very hot in my area, the room is around 30°C, and I just noticed that my roughly two year old hAP ax³ reaches about 72°C in idle and 83/85°C when the CPU is at 50% for 5 minutes.
right, stock thermal paste dried out too fast and stock thermal pads didn’t cover the full surface. The PCB design allowed to put chunky aluminum radiator yet not attached as needed. Thermal pad’s thickness was incorrect too, I confirmed it with forming a ball shape glue tack and tightened the radiator’s screws.
I did reapply thermal paste and pads three times, second time intentionally covering only part of the contact surface, that second time proved that high packet loss and excessive temperature is nothing but result of thermal dissipation problem. CPU temperature was too high even with lower CPU clock. In my experiments the load and CPU temperature didn’t correlate as expected. The load didn’t contribute much unless I ran load generator. Temperature was either medium (first, third time) or very high (second time).
Third time I applied thermal paste and pads as even as possible, checked thickness and surface area, then packet loss was gone and CPU temperature remains within +50C (few degrees higher during past summer).
Maybe it is also worth mentioning that CPU temp sensors are usually directly integrated into the dye (chip) inside the case.
60°C sound worrying if one imagine a chip-package or case surface. Inside a chip, this is not a problem. About 30° delta to env temp is to be expected for passive cooled devices.
Modern CPUs can go up to 100°C chip (junction) temp. And even this limit is not because of the heat would burn something, but the thermal motion of semiconductor atoms starts to push out dotation atoms what destroys the chip's functionality.
