I capture an SD image from my RPi2 to my file server each night. While this occurs there is an increase in the CPU temperature. Curiously, there is a decrease in CPU utilization shown on the collectd graphs. This inverse relationship distinctly occurs whenever there is some other intensive CPU activity such as a long package upgrade. I’m sure it is just an artifact of the data being plotted but on its face, it is counter-intuitive. Anyone else seen this behavior? I suspect it is an artifact of the data, but anyone know why?
IIRC recent raspbians configure cpu frequency scaling so it may be that your image capture is enough to trigger the cpu to speed up. In that case the decrease in the graph isn’t actually a decrease in CPU used, it’s an increase in CPU capacity.
Interesting; makes sense. I was trying to pin it to the multi-cores somehow but wasn’t make the connection. Thanks.
I see exactly the same relationship with exactly the same activity - nightly image backup.
I do also see USB CPU use increase, which is no surprise in my case as I’m writing the image directly to a network share.
Keep in mind that your CPU chart is only showing the utilization of the dump1090 processes. If you look at an overall CPU chart, you will note an increase in both utilization and temperature at the same time. Here’s an example showing both charts.