Hello, I am trying to test the jetson nano when doing a continuous work/task with GPU (neuronal network, exactly yolov5s in torch 1.8.0 in python3).
In a previous questions, I manage to broke my jetson (at least I guest that), and evething points to temperature. I bought another and boot the sd card and then tried the sd card in the “broken” jetson nano and it works, so i think its still “operational”. Anyways,… now I want to test the task controling the temperature (since I think was the problem with the last nano).
This “task” will be performed for 12-14 hours continuously, environment temperature might be tricky since it will be placed somewhere with aircondition but people could turn it off.
So I checked this commands
$ cat /sys/devices/virtual/thermal/thermal_zone*/temp
57000
50000
49000
45500
100000
49750
$ cat /sys/devices/virtual/thermal/thermal_zone*/type
AO-therm
CPU-therm
GPU-therm
PLL-therm
PMIC-Die
thermal-fan-est
As you can see they give some information about the sensor temperatures. Even though AO,PLL,PMIC-Die and them-fan are not completely understood.
So, what those abrevations means?, and more important, what they refer to? Does AO is ambient temperatura sensed by the jetson nano?
Also, what would be a good range to keep all this temperatures ok? I see documentation about CPU and GPU but I do not really know if i should take care of the others?
So, for example, lets say I am doing my task in the nano, and this makes the nano goes to AO:75000,CPU70000,GPU71000 (just using those 3 for the example). Should I turn the task off until the nano returns to a better temperature? Something like that would be great, because as far as I can relate, the nano will not kill the task, but it will reboot.
Thanks in advance for any help.
UPDATE: when performing the task, AO goes up to 60C+, and jtop show it with a red color.
Note: This jetson nano have no fan. Is the dev kit as packed, just the normal heat sink.