No, it's on a much smaller timescale. This is 20 minutes on a BitAxe, that's already a 10% difference, and it keeps going up and down. Temperature and voltages look completely stable. It must be something about the calculation.
Although what I did notice is that the measuring points are sometimes further apart than other times, so maybe it's something about the template distribution? Lots of tx's on the top of the fee range coming in makes a lot of new templates, which makes it slow down?
