Lets forget the influence that the DMC sampler has for now and say it’s not adaptive. I just want to understand the basic concepts. And doesn’t 1 subdiv = 4 samples?
Or would it only fire 1 subdiv (4 samples) hemispherically from the interstection of each eye ray from the camera perspective? or would it be from every scene polygon - so lighting calcs are still carried out for objects outside of the frame (not talking about GI)
EDIT - note, I was meant to quote both Joconnell and DADALs response.