In addition to rendering process, they are also suitable to general compute-intensive, highly parallel computation. NVIDIA’s GPU with the CUDA programming model provides an adequate API for non-graphics applications. CPU sees a CUDA device as a many-core co-processor.