Well, that’s the thing. No new threads are created with Nexus, it adopts cooperative multithreading, which means that a dynamic number of threads is in the thread pool at any given moment, depending on how many tasks you schedule. The maximum number of threads is a command line argument and defaults to the number of cores * 2.
The perfect way to leverage the whole hardware is to exploit — kernel-assisted — multi-threading, not abandon it for an IPC model that adds overhead, and If we’re talking about job queuing, then cooperative multitasking is superior to any other model I’ve seen so far.
Time alone will prove which is the better model.