Numpy mean of large array slower than mean of mean of all axes

Question

Running Numpy version 1.19.2, I get a better performance doing the mean of an array by cumulating the mean of every individual axis than by calculating the mean over an alre