This approximate solution to the same (NP-hard) MAX-CUT problem for a N = 2000 complete graph is competitive with a modern algorithm in the computation time taken to select a solution that can guarantee an 87.8% approximation accuracy.
On installing a classical neural network called a Hopfield-Tank in a RIKEN supercomputer as an algorithm and computing an approximate solution, the time taken to arrive at this solution was 7 ms, but that by the NTT CIM was only 70 μs, about 100 times faster.