You can just run a few devices through a power monitor. Watch them at idle for a baseline, and then run a system stress - unless what you're saying is the benchmark itself is incomplete? And adding, this technique works better on realtime systems because you won't have seemingly random variances in resource usage like you will on a windows system for example.

Check out my IoT graphics library here: https://honeythecodewitch.com/gfx And my IoT UI/User Experience library here: https://honeythecodewitch.com/uix

Daniel Pfeffer wrote:

Your suggested method is good for "black box" testing, but won't separate application vs kernel power usage, for example. If one really wants to measure the application's power usage, one should separate out the kernel's power usage, to say nothing of other applications running simultaneously.

That's why I suggested a realtime system would give better results. You don't really have much in the way of context switches and other kernel overhead, or if you do, you can absolutely control it.

Check out my IoT graphics library here: https://honeythecodewitch.com/gfx And my IoT UI/User Experience library here: https://honeythecodewitch.com/uix