previous | contents | next

752 Part 3 ½ Computer Classes Section 4½ Maxicomputers
 
 

General Eclipse computer in this category. The Cray Research "A" processor, a 16-bit, 80 MIPS minicomputer is scheduled to replace the Eclipse in early 1978. Front-end computers can be attached to any of the CRAY-1's 12 i/o channels.

The physical connection between a front-end computer and the CRAY-1 is shown in Fig. 8. In this example an IBM 370/168 is assumed in the front-end role. Note that each computer requires a channel adapter between its own channel and a Cray Research phase-modulated long line. The link can only be driven at the speed of its slowest component. In this example it is the IBM block multiplexer channel speed of 3 megabytes/second. The discipline of the link is governed by the Cray Link Interface Protocol.
 

CRAY-1 Development Problems1

Two of the most significant problems encountered on the way to the CRAY-1 were building the first cold bar and designing circuits with a completely balanced dynamic load.
 

Building the Cold Bar

It took a year and a half of trial and error before the first good cold bar was built. The work was done by a small Minnesota company. A major problem was the discovery, quite early, that aluminum castings are porous. If there is a crack in the stainless steel tubing at the bond between the tubing and the elbow then the Freon leaks through the aluminum casing. The loss of the Freon is not itself a problem, but mixed with the Freon is a little oil, and the oil can cause problems if it is deposited on the modules. Aluminum also tends to get bubbles in it when it is cast, requiring a long process of temperature cycling, preheating of the stainless steel tube, and so on.

Designing the Circuits

CRAY-1 modules are 6 inches wide. The distance across the board is about a nanosecond which is just about the edge time of the electrical signals. Unless due precautions are taken, when electric signals run around a board, standing waves can be induced in the ground plane. Part of the solution is to make all signal paths in the machine the same length. This is done by padding out paths with foil runs and integrated circuit packages. All told, between 10 and 20 per cent of the IC packages in the machine are there simply to pad out a signal line. The other part of the solution was to use only simple gates and make sure that both sides of every gate are always terminated. This means that there is no dynamic component presented to the power supply. This is the principal reason why simple gates are used in the CRAY-1. If a more complex integrated circuit package is used, it is impossible to terminate both sides of every gate. So all of the CRAY-1's circuits are perfectly balanced. Five layer boards have one ground layer, two voltage layers, and then the two logic layers on the outside. Twisted pairs which interconnect the modules are balanced and there arc equal and opposite signals on both sides of the pairs. The final result is that there is just a purely resistive load to the power supply!
 

Summary

The design of the CRAY-1 stems from user experience with first generation vector processors and is to some extent, evolved from the 7600 Auerbach [n.d.]. The CRAY-1 is particularly effective at processing short vectors. Its architecture exhibits a balanced approach to both scalar and vector processing. In Keller [1976], the conclusion is drawn that the CRAY-1 in scalar mode is more than twice as fast as the CDC 7600. Such good scalar performance is required in what is often an unvectorizable world.

At the time of this writing, Cray Research has shipped CRAY-1 systems to three customers (Los Alamos Scientific Laboratory, National Center for Atmospheric Research, and the European Center for Medium Range Weather Forecasts) and has contracts to supply three more systems, two to the Department of Defense, and one to United Computing Systems (UCS). Production plans already anticipate shipping one CRAY-1 per quarter. As the population of CRAY-1 computers expands, it will become clear that the CRAY-1 has made a significant step on the way to the general-purpose computers in the future.
 

References

Auerbach [n.d.]; Calahan, Joy, and Orbits [n.d.]; Computer World [1976]; Keller [1976]; O'Murphy and Wade [1970]; Reeves [1975].
 
 

1This section on CRAY-1 development problems is based on remarks made by Seymour Cray in a speech to prospective CRAY-1 users in 1975.
 
 

previous | contents | next