another status update: we just started a 4-player direct ip pitboss game tonight (4 humans on team 1, 3 additional teams of 4 AI players each, huge map, permananent alliances, simultaneous turns).
of the 4 human players, 3 were on the internet, and 1 was on the same network as the pitboss, behind a NAT router.
we had no network connectivity issues, but compared to last night's game with 4 teams of 3 civs each (3 humans, 9 AIs, huge/permanent/simultaneous), performance was *horrible*. delay between turns was extreme, and pitboss was using 30% cpu consistently, on an amd athlon 64 2.2ghz with 2GB memory. going from 12 total civs (3+9) to 16 total (4 human + 12 AI) really seems to have killed performance in a big way.
or maybe we just got unlucky. a sample size of 1 is not statistically significant, after all.
-ken