Overheating on mainline Linux

[Roman]

With the current Linux-libre, I observe that A64-OLinuXino may overheat under heavy prolonged load.

Interestingly, when I try "stress-ng -c 0", which fully loads all four cores of the CPU, the temperature stabilizes right above the default lower temperature trip point of 75 degrees, even without a heatsink and in BOX-A64-BLACK inside an unventilated cabinet. But that load is synthetic and even.

Under a real-world load, for example, when I run "guix pull", which produces an uneven load over a long period of time (probably an hour), the temperature drifts and at some point may exceed 110 degrees. The last message that I see in the serial console says that 110 degrees were reached and the device is going into reboot. The reboot probably happens too fast for the chip to cool down because the board never starts by itself.

I conclude that the temperature fluctuates too fast under an uneven high load. The default configuration of the kernel does not handle that well. The configuration could probably be changed by configuring the power allocator. But I think that adding even a small heatsink on the A64 BGA should completely resolve the issue. The heatsink will be absorbing heat and should smooth the temperature spikes.

Olimex lists ALUMINIUM-HEATSINK-20x20x6MM as a related product for Allwinner A64. I suppose that it is appropriate for the task. I also saw a member of the forum reporting that a 15x15x15 is the right size for A64 and fits perfectly inside BOX-A64-BLACK. A datasheet for A64 says that it comes in "15x15mm" FBGA396 package. So, I probably should be looking for a 15x15 heatsink. Most of the readily available radiators are 14x14 mm in my local market, however.

If anyone has any advice on preventing overheating or on choosing a heatsink, it will be much appreciated.

Code Select Expand

$ uname -r
7.0.12-arm64-generic
$ cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor | uniq
schedutil
$ ls -l /sys/class/thermal/thermal_zone*/cdev*
lrwxrwxrwx 1 root root    0 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev0 -> ../cooling_device0
-r--r--r-- 1 root root 4096 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev0_trip_point
-rw-r--r-- 1 root root 4096 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev0_weight
lrwxrwxrwx 1 root root    0 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev1 -> ../cooling_device0
-r--r--r-- 1 root root 4096 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev1_trip_point
-rw-r--r-- 1 root root 4096 Jun 20 12:18 /sys/class/thermal/thermal_zone0/cdev1_weight
$ cat /sys/class/thermal/cooling_device0/type
cpufreq-cpu0
$ cat /sys/class/thermal/thermal_zone0/trip_point_1_type
hot
$ cat /sys/class/thermal/thermal_zone0/trip_point_0_type
passive
$ cat /sys/class/thermal/thermal_zone0/trip_point_0_temp
75000
$ cat /sys/class/thermal/thermal_zone0/cdev0_trip_point
0
$ cat /sys/class/thermal/thermal_zone0/cdev1_trip_point
1
$ cat /sys/class/thermal/thermal_zone0/trip_point_0_hyst
2000
$ cat /sys/class/thermal/thermal_zone0/k_po
0
$ cat /sys/class/thermal/thermal_zone0/k_d
0
$ cat /sys/class/thermal/thermal_zone0/sustainable_power
0