Since the GPU has a high parallel structure, the GPU with data in the image processing complex algorithms and CPU efficiency higher than that of the CPU .GPU there are differences in the structure, most of the area for the controller and CPU registers, compared more GPU with the ALU (arithmetic logic unit) for data processing, and data cache instead of flow control, such a configuration is suitable for parallel processing of data-intensive .CPU at execution time, a time only one data processing, there is no real meaning in parallel, while the GPU core having a plurality of processors, data can be processed in parallel at a time.
GPU using parallel computing streaming mode, can be independently calculated for each of the parallel data, the "data calculated independently", i.e., any element within the flow calculation does not depend on the same data with other types, e.g., calculating a vertex world coordinate position, does not depend on the position of the other vertex. the so-called "parallel computing" means "may be used simultaneously a plurality of data, the time data of a plurality of parallel computing time and a separate data execution is the same."