Reader Question: Cell Does Graphics?
Mike has another question:
In regards to the “Multiple PS3â€™s Working Together” question; how is it that the cell processor can produce graphics, especially graphics that reach the level of the e3 05′ London demo? I thought that the cell processor only made large mathmatical calculations and nothing more. If the cell processor can act as a gpu in certain ways and work together with the nVidia gpu, then ps3 game graphics really will br amazing.
What is a GPU except a special-purpose processor? A GPU is basically a CPU but will special circuitry for doing 3D graphical things faster than a normal CPU would. A general-purpose CPU like the Cell is perfectly capable of rendering awesome 3D imagery. The faster it is, the better the imagery. And the Cell is fast. Also, it has 7 SPE’s. 3D images are easy to partition into pieces, which is great for a multi-core processor like the Cell.
Witness RenderMan and MentalRay. These two renderers (while not realtime) produce the best looking CG imagery in the industry today, and run on normal PC’s using the main CPU. Gelato by NVIDIA is similar, except it accelerates the rendering using NVIDIA hardware. So Gelato uses both the main CPU and the GPU. Many games use mostly the GPU, just feeding it data (though the CPU still needs to do some of the setup work). So if you want to create a 3D image, you can do it with a CPU alone, CPU and GPU together, or mostly the GPU. Photorealistic imagery is too compute-intensive to not use the CPU, and even then takes many minutes to render a frame.
However, if you’re willing to sacrifice Finding Nemo-type graphics to get realtime interactivity, you have options. You can use the CPU to render a scene, (mostly) the GPU to render a scene, or a combination thereof. It’s really up to the game developer what they want to do. It just so happens that the Cell is so powerful, it does a great job all by itself, even without that special 3D circuitry.