The answer to this question depends on the resources required by your code.
If your code can use multiple processors on the same node or across multiple nodes, you should see a significant speed up in processing time.
However if your code can only run on a single processor, you may be able to queue up multiple single processor jobs, thereby reducing the time it takes to process your data.
Finally GPU optimized code can give significant performance boosts when using CUDA and our Tesla cards.
Unfortunately there are no hard and fast rules, generally it is taken on a case by case basis.