I wouldn't say "better"; I'd probably say that the memory profile would be completely different. It all depends on the number of actors, the size of objects, a

I wouldn't say "better"; I'd probably say that the memory profile would be completely different. It all depends on the number of actors, the size of objects, and the workload. Remember: for large binary objects, references on a shared heap are sent and messages are not explicitly copied per actor. This means it's dependent on the workload and both the structure of how many processes are involved in a single request.