What is the faster way to find the maximum along axis on a symetrical matrix in numpy

Question

Suppose I have a symetrical numpy array as follows:

import numpy as np

N = 10000
b = np.random.random_integers(-2000,2