Why exactly does numpy.sum() gives different results on the same matrix with different order, where sum() does not?

Question

When I initialize two random matrices (here of shape (6,2)), and re-order the rows, np.sum() will give me different results. I understand this is likely due to numerical errors,