Supercomputing is essential for many fields such as theoretical physics, engineering, and even precision medicine. I will explain the challenges involved in designing a supercomputer that achieves high sustained performance on many thousands of nodes, in particular the balance between processor, memory system, and network. I will describe several successful examples, including some I have helped to design in collaboration with IBM and Intel. I will also discuss very recent developments based on ARM technology, such as the Post-K supercomputer in Japan.

Tilo Wettig obtained his PhD in theoretical physics from Stony Brook University in 1994and then did two postdocs at the Max-Planck-Institute in Heidelberg and at the TechnicalUniversity of Munich. He joined the faculty ofYale University in 1999, with a jointappointment at the RIKEN Brookhaven ResearchCenter. In 2004 he moved to the University ofRegensburg, where he holds a chair in theoreticalphysics.