I can't recall ever having a NS performance problem caused by code execution speed. Invariably it's some form of IO at fault in all cases.