What's the point of multiplexing tasks on a particular core if the tasks don't do any I/O? It will be strictly faster to execute the tasks serially across as many cores as possible then.