North Carolina State University claims to offer just that. Despite the emergence of fused architecture SoCs, the CPU and GPU cores typically still work independently. The University hoped that by assigning tasks based on each processor's abilities, performance efficiency would be increased. As the CPU and GPU can fetch data at comparable speeds, the researchers set the GPUs to execute the computational functions, while the CPUs did the prefetching. With that data ready in advance, the graphics processor unit has more resources free, yielding an average performance boost of 21.4 percent though it's unclear what metrics the researchers were using. Incidentally, the research was funded by AMD, so no prizes for guessing which chips we might see using the technique first.