The latest revision is a completely different test. As klodolph explains in another comment, the presumably common case of no exception is generally how try/catch is optimized. So it stands to reason that the exception case would be slower and that there is potential for the non-exceptional case to be faster.