Comparison of features for current commercial multicore

Published by the IEEE Computer Society 0018-9162/10/$26.00 © 2010 IEEE In the past, developers used additional capacity to develop superscalar CPUs with replicated execution units and deep pipelines to exploit instruction-level parallelism. However, they only harvested about 25 percent of the additional chip space that became available per year by adding new architectural features.2 Moreover, the performance gap between processors and memory limits the gains possible from further increasing processor frequency. Thus, the design direction currently employed for performance increases uses available chip space for multithreaded and multicore CPUs. These designs support multitasking via parallel programs or running several applications concurrently. Designers first introduced multithreaded CPUs, which employ hardware-level context switching between threads, to reduce the idle time of resources in complex superscalar processors. Shortly after this, designers integrated more than one processor core onto a single chip, and we now have eight-core processors. Assuming that Moore’s law holds, we expect a doubling of the number of cores on chip every two years, leading to CPUs of 16 or more cores in the near future.