System architecture for large-scale integration

Abstract

The developing capability of the semiconductor industry to fabricate and interconnect a hundred or more logic gates on a single silicon chip promises to have a substantial impact upon the performance and reliability of today's computers. In just a decade, computer fabrication techniques have progressed from a single vacuum tube gate occupying many cubic inches in volume, to second generation discrete transistor circuitry, and to integrated circuit flat-packs in the third generation machines. Each successive generation has offered more computing power through faster circuitry and increased packing densities. Approximately 99% of the volume, even in densely packaged third generation computers, represents packaging and circuit interconnection material, and this separation between computer components still represents a severe speed bottleneck. It is not uncommon for 75% of the machine delay to occur in interconnection wiring with only 25% of the delay inherent in the flat-packs. Large-scale integration of logic gates on a single silicon chip offers promise of breaking this speed bottleneck in the larger and faster fourth generation machines.