Mistershine said:
48 cores? Do each of these cores need to be accessed individually or can devs just throw code at the GPU and let it sort it out itself? |
Both (if everything goes to plan). Intel will have a standard DirectX/OpenGL driver which will 'look' like a normal GPU to a game and that will divide it up to the cores, OR you can program each one individually like a CPU. I think they'll also have their own special API like CUDA or Stream where the task splitting is semi-automated.







