HPC system architectures are shifting from the traditional clusters of homogeneous nodes to clusters of heterogeneous nodes and accelerators. The future of high-performance computing (HPC) from the technologies developed today to showcase the leadership-class compute systems, the supercomputers. These machines are usually designed to achieve the highest possible performance in terms of the number of 64-bit floating-point operations per second (flops). Their architecture has evolved from early custom design systems to the current clusters of commodity multisocket, multicore systems.
No takes yet. Share an insight, caveat, or question.
Kindratenko et al. (2011) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: