Guest Editors’ Introduction: Hot Chips II Symposium

David A. Wood q Universfty of Vhconsinhe unique Hot Chips Symposia attract presentations by the designers of the most exciting recently developed chips. By bringing together many of the field’s best academics, industrial researchers, and product developers, these conferences provide a dynamic forum for exchanging new ideas. We invited authors of the best of the 26 presentations at the Hot Chips II Symposium to submit papers for this special issue. From those submissions we selected the five articles you see here. A theme running through many of the presentations is that superior chips require advances in both technology and organization. Technological advances improve the speed of the underlying circuitry, thus reducing the latency of each operation. Despite rapid technological advances. however, merely riding the technology cume proves insufficient. Organizational advances must also take place to allow more hardware components to operate in parallel, thereby increasing the rate at which operations complete (throughput). The highest performance chips combine reduced latency from advanced technologies with increased throughput from parallelism. There are nvo ways to exploit parallelism in computer systems: change onl), hardware or change both hardware and sofnK]re. The first evolutionary approach requires predictable risks in exchange for predictable gains. The second approach. a revolutionary one. invol\-es suhstan-