Steve Vinoski improves upon Tim Bray’s code and tries out different configurations:

As we increase the number of Erlang processes, our performance improves, up to a point. At 40,000 processes we’re slower than we were at 4000. Maybe there’s a better number in between? It turns out that despite the numbers listed above, once you get above 400 processes or so, the numbers remain about the same. The best I got on my MacBook Pro after numerous runs was 0.301 seconds with 2400 processes, but the average best seems to be about 0.318 seconds.