I your code can fit completely within the memory cache within the processor you can speed up execution quite a bit. Granted, it's not something you'd be really worried about in a business app.

There's one story I recall back from my university days -- where a simple change in code (changing array indicies around) made a signficant difference in program execution. The program in question was written by someone who wasn't paying attention (nor was aware) of how arrays were implemented in the particular language (e.g. row major or column major). In a nutshell due to the way the program was written it was generating a page fault on almost all array accesses. A simple swap of array indicies significantly reduced the number of page faults and allowed the program to actually complete execution (previously it would never finish and would get kicked out when they brought the computer down for PM).