Cray-2

Cray-2
Name	Cray-2
Developer	Cray Research
Designer	Seymour Cray
Introduced	1985
Discontinued	1990s
Cpu	4 vector processors
Memory	up to 2 gigabytes
Predecessor	Cray-1
Successor	Cray Y-MP
Type	Supercomputer

Cray-2

The Cray-2 was a vector supercomputer developed in the mid-1980s by Cray Research under lead designer Seymour Cray, intended to succeed the Cray-1 and to push the boundaries of high-performance computing for agencies such as the Los Alamos National Laboratory, Lawrence Livermore National Laboratory, and corporations like Boeing and Lockheed. It combined aggressive circuit design, three-dimensional packaging, and innovative cooling to deliver peak performance aimed at demanding workloads in computational fluid dynamics, nuclear weapons simulation, and weather modeling used by institutions including NASA and Naval Research Laboratory. The system’s deployment influenced procurement at Electric Power Research Institute and research at universities such as University of Illinois Urbana–Champaign and Stanford University.

Development and design

The Cray-2 project originated after the commercial success of the Cray-1 and was motivated by requirements from customers like Department of Energy laboratories. Development centered at Chippewa Falls, Wisconsin and the Cray Research headquarters in Sioux Falls, South Dakota, with design leadership from Seymour Cray and engineering contributions from teams with backgrounds at Control Data Corporation and Applied Research Corporation. Management interacted with procurement officers from Los Alamos National Laboratory and program managers at Lawrence Livermore National Laboratory to define goals for sustained vector throughput, memory capacity, and reliability. Budgetary and scheduling pressures mirrored earlier programs such as ARPANET technology transfers and later procurement debates like those surrounding the Cray-3.

Architecture and hardware

The machine featured multiple vector pipelines organized across four identical processor modules derived from prior work at Cray Research and concepts traced to projects at Bell Labs and IBM Research. Each processor implemented vector registers, scalar registers, and a sophisticated scoreboard inspired by designs at CDC and I.B.M. laboratories. A high-bandwidth memory subsystem used static RAM chips supplied by vendors including Intel and Texas Instruments, arranged to provide capacities up to two gigabytes and bandwidths aimed at streaming applications found in computational fluid dynamics codes developed at MIT and Caltech. Interconnect and I/O subsystems supported peripherals from Fujitsu and interfaces compatible with software environments prevalent at University of California, Berkeley and Carnegie Mellon University.

Performance and benchmarks

Cray-2’s advertised peak performance targeted several hundred megaflops to a gigaflop range depending on configuration, evaluated using benchmarks influenced by the practices at SPEC and internal tests similar to workloads at Los Alamos National Laboratory. Real-world performance on vectorized codes in turbulence simulation and nuclear modeling—areas pursued by Lawrence Livermore National Laboratory and Sandia National Laboratories—showed substantial speedups over the Cray-1 for memory-bound problems but mixed gains on scalar-heavy workloads encountered at General Electric research labs. Comparative studies published in venues frequented by researchers from Stanford University and Princeton University influenced procurement choices at institutions like National Center for Atmospheric Research.

Cooling and packaging

A hallmark of the design was immersion cooling using a fluorocarbon dielectric provided by firms such as 3M; the system’s sealed tanks and three-dimensional packaging echoed thermal approaches experimented at Hewlett-Packard and in aerospace electronics at Grumman. The close packing of circuits and use of multiple layers of boards were informed by packaging research at MIT Lincoln Laboratory and by high-density assembly techniques developed with partners in Japan and West Germany. Cooling enabled high clock rates but introduced operational challenges that required facilities support by organizations like Argonne National Laboratory and maintenance agreements with Cray Research service engineers.

Operating system and software

Software environments for the machine included a proprietary operating system developed by Cray Research that supported batch scheduling and vectorizing compilers originating from research at University of Illinois Urbana–Champaign and University of California, Berkeley. Fortran compilers and libraries were optimized for vector instructions following techniques matured at Lawrence Livermore National Laboratory and echoed in later work at Oak Ridge National Laboratory. Application codes for computational chemistry, structural analysis, and climate modeling were ported from projects at Caltech, Scripps Institution of Oceanography, and corporate research groups at Boeing.

Deployments and users

Major installations were at Los Alamos National Laboratory, Lawrence Livermore National Laboratory, and National Center for Atmospheric Research, with commercial users including Boeing and Lockheed. Academic sites such as Stanford University, Princeton University, and University of Illinois Urbana–Champaign hosted Cray-2 systems for research in numerical linear algebra, computational fluid dynamics, and parallel algorithm development. International customers included organizations in United Kingdom, France, and Japan, reflecting global demand from scientific institutions like CERN and national laboratories comparable to Brookhaven National Laboratory.

Legacy and impact

The machine’s innovations in packaging, immersion cooling, and vector processing influenced successor systems such as the Cray Y-MP and later architectures explored at Silicon Graphics and Convex Computer Corporation. Lessons from its operational experience informed procurement policies at Department of Energy laboratories and contributed to academic research in high-performance computing at MIT and Carnegie Mellon University. The Cray-2’s design choices echoed in later supercomputing projects supported by initiatives like National Science Foundation programs and helped shape the evolution of vectorization, cooling technology, and high-density packaging used in exascale efforts at facilities such as Oak Ridge National Laboratory and Argonne National Laboratory.

Category:Supercomputers