General GPU operation

As their name suggests, GPUs are primarily designed to run the graphics pipeline. This involves the simultaneous processing of a large number of vertices and an even greater number of pixels. For example, real-time rendering requires the processing of billions of pixels per second. This implies massive, fine-grained data parallelism, with mostly the same processing on each pixel. This leads to an execution model that is fundamentally different from that of general-purpose processors (CPUs). Figure 6 shows the difference between the two models: