I could theoretically split the table up, but I would (from a pure data size view) need to duplicate at least the primary key, and I wouldn't really gain anything in overall size or size of indexes, ...