The inverse of your statement is also true. Denormalizing the database to avoid duplicating data will not necessarily improve performance in an RDBMS - it might even make it worse.