Some memory-touching orders are better than others, because CPUs will prefetch nearby memory into CPU cache during memory accesses.