Depending on the processor, it can also be the difference between instructions in a tight loop only needing to be decoded once (because of micro-op caching) vs. decoded each loop iteration.