Evaluating a Multithreaded Superscalar Microprocessor versus a Multiprocessor Chip

This paper examines implementation techniques for future generations of microprocessors. While the wide superscalar approach, which issues 8 and more instructions per cycle from a single thread, fails to yield a satisfying performance, its combination with techniques that utilize more coarse-grained parallelism is very promising. These techniques are multithreading and multiprocessing. Multi-threaded superscalar permits several threads to issue instructions to the execution units of a wide superscalar processor in a single cycle. Multiprocessing integrates two or more superscalar processors on a single chip. Our results show that the 8-threaded 8-issue superscalar processor reaches a performance of 4.19 executed instructions per cycle. Using the same number of threads, the multiprocessor chip reaches a higher throughput than the multithreaded superscalar approach. However , if chip costs are taken into consideration, a 4-threaded 4-issue superscalar processor outperforms a multiprocessor chip built from single-threaded processors by a factor of 1.8 in performance/cost relation.