LinHT – SoM thermal management

Recently, Jacek SQ5BPF performed a very nice experiment. The question that we had was this: how does the SoM behave, thermal-wise, under heavy CPU loads? Is any form of cooling required?

Jacek plotted a temperature vs. time graph after logging its values for a while, using a simple bash script. It read the CPU’s thermal sensor at a fixed interval, while the machine was stressed with sysbench1:

#!/bin/bash
#just get the temp from linht --sq5bpf
while : ; do
echo -n "$SECONDS  "
echo "scale=2;`cat /sys/devices/virtual/thermal/thermal_zone0/temp`/1000" | bc
sleep 1

The values were used as-is (the sensor is not calibrated). CPU stress was relieved after 10 minutes. Two test setups were examined: without thermal pads and with pads applied. Thermal conductivity reported by the manufacturer is 6 W/m⋅K.

1.5mm pads, shaped to even up the bottom of the chassis with the protrusion.
2mm pads added on top.

The results are shown below.

CPU stress was stopped at around 600 seconds mark.

The CPU clearly runs at much lower temperatures with the pads applied. The test has been conducted at ambient temperature of about 22°C. The CPU idles at 36.5°C with pads applied, which is over 5°C cooler than the baseline.

Thank you a lot for performing this experiment!

  1. sysbench --time=600 --threads=2 cpu run ↩︎

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *