Numpy significantly slower on Windows than on Ubuntu

Question

I ran the following snippet in IPython on both Windows 10 and Ubuntu 16.04 installed on the same machine (dual-boot):

import numpy as np

a = np.random.random((1280, 1024))
b = np.random.random((1280, 1024))

%timeit a / b

On Windows, this outputs:

1.81 ms ± 21.3 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

On Ubuntu, the output is:

649 µs ± 1.41 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

I see the same performance difference with all basic arithmetic operations.

On both systems, I'm running this in a conda environment, with Python 3.8.3, NumPy 1.19.2 and MKL 2020.1. The machine runs AMD Ryzen 9 5950X CPU.

Why is there such a big discrepancy between the two systems? Are there ways to bring performance on Windows closer to Ubuntu?