TDP is assumed at 50%. Why?

[flyaruu]

Hi,

I came across this assumption, around this piece of code:

// taking the midpoint of the two datapoints and dividing by 50 because we're
// assuming tdp is at 50% utilization
(0.5 * (x.cpu_usage + y.cpu_usage)) / 50.0 * tdp * delta_t_h

Is there an official source here? I've tried it locally on a mini pc, and there TDP seems to roughly align to 100% usage.
(Tdp is 15w, max power consumption of the entire PC is about 21w. At 50% it is exactly 15w, but I have no way to know how much of that is the CPU)

It's very influential on the output figures, so the ratio should be documented somewhere.

[ohuu]

TDP probably shouldn't be used at all because we can create a power curve if we know the TDP. We should generate a power curve by calculating a, b, c and d given the overall TDP and use the power curve instead