Cray-3

The supercomputer Cray Cray -3 is the successor of the Cray -2. The system should be the first application of gallium arsenide semiconductors ( GaAs semiconductor called also ) be in electronic data processing, which had already been tried in the previous model. However, the project was a huge flop because the Cray -3 due to its gallium arsenide circuits was very expensive and therefore only one copy was delivered.

History

Immediately after the completion of the Cray 2, Cray Seymour made to the development of the Cray - third This calculator he left Cray Research and founded his third company Cray Computer Corporation ( CCC ). It was a peaceful divergence. The reason: the two-year delay in the Cray -2, which should appear in 1982, it was necessary to establish a successor to the Cray -1 - finally bringing the competition out new models. The choice fell on a development of the Cray -1 on two processors, the Cray X -MP. It was possible to develop them further (later it was the Cray Y -MP with 4 or 8 processors and Cray T90 with 16 processors ), but no company can afford to practically drive two supercomputers lines: Such computers are sold individually and from a model rarely more than 100 pieces.

The problem that already delayed the development of the Cray -2, the cycle times were: Slowly we came to borders that were imposed by the physics. The light and thus also the clock signals move away in a nanosecond only 30 cm. Transistors switch in areas which, although they are still below 1 ns, but in the processing of data, the data is passed through a few dozen gates, so that adding the switching times.

Seymour Cray saw the solution in another semiconductor material: gallium arsenide. Although silicon is the material most commonly used and also the cheapest. But it is the worst of the electrical properties. Both germanium transistors and the semiconductor elements of the third and fifth main group, such as gallium arsenide switch much faster, because the electrons have a higher mobility. Germanium comes on price and practical considerations for highly integrated circuits out of the question, yet remained gallium arsenide.

What Seymour Cray has underestimated was the complexity of the task: To develop a new computer is already difficult. But to do to construct new circuits on the basis of a material not commonly used, was too difficult. The development was delayed from 1988/1989 to 1993. Backed by an order and government support, he could imagine at a cost of 120 million USD in 1993 a processor of the Cray 3 and install it into the NCAR. The entire system should consist of 16 processors consist in a mere 90 × 90 cm cube, 4 GB memory and direct connections with 8 GB / s between processors. With a cycle time of 2.11 ns, a processor achieved a peak performance of 0.948 GFLOP, the entire computer GFLOP 15.17. The power consumption of the small cube was 88 kW.

But after the Cray -3 was ready for production, they have no one wanted - it was just too expensive. Similar thing has happened to the CM -2 Convex, another supercomputer manufacturer, the gallium arsenide began.

Cray Research was the architecture of the Cray -1 and further developed with the Cray C90, a 16- processor system with a performance of one GFLOP per processor - only 4.2 ns cycle time and conventional technology. In 1995, Cray's company CCC bankrupt. She can not sell a Cray - 3 system. In the meantime, however, Seymour Cray already worked on the Cray -4 - also with gallium arsenide and 1 ns access time. On September 22, 1996, he died following a traffic accident before he could complete the Cray -4.

System Information

Supercomputer

Cray-2 Cray-1 Cray X-MP National Center for Atmospheric Research Cray-4

206254