[ZirconiumX]

My opinion is that this post reaches the same conclusion as I would, but I disagree with the logic behind it.

In a modern, state of the art chess program that uses alpha-beta with a branching factor b and depth d, you will have at most d boards allocated, or even 1 board allocated. Neither of those figures approach b^d. That makes board size largely irrelevant, except for the amount of memory copying needed, which for a d-board approach would be b^d copies (a single board approach would only mutate that board).

EDIT: One major issue I just noticed with the article is that the two differing implementations are not apples-to-apples equivalent. One uses bitboards, based around manipulation of bits, and another one is akin to representing the board with an array.