The four threads solution only takes the two threads and splits each into two. In a dual cpu, each gets assigned its own region, so in four threads in Poser, each core is rendering the same thing as in two threads, just splitting the task in half. It does in fact go "side to side" on each of the two spilt processes.
I'm not seeing any increased speed or any reason to keep it set like that, but it does look a little more interesting...