Non Sequor said:
A GPU has vastly more parallelism than the Cell does. Generally speaking, it's not going to make sense to offload tasks from the GPU to the CPU. |
.. i didnt read a 600 page tech document on the cell released by IBM and not know what im talking about... the Cell is completely designed for parallel processing. Even though it could slow down tasks from GPU to the CELL you can greatly speed up tasks going CELL to GPU. Each SPE can run 3 tasks (but can be forced to do 4 if they are small tasks) so you can have physics calculations running on an SPE and unload it to the GPU to render what was calculated speeding up the rendering time.








but don't they have to work with each other? Excuse me if i'm wrong, i'm not up to date with this


