Cray and NEC - an unbelievable science fiction story - the past

UCAR is an affiliate of the federally funded National Science Foundation. The whole story started in July 1996, when Cray filed its complaint with the Commerce Department. NEC won in May 1996 the UCAR tender. A NEC SX-4 better fitted the needs and performance requirements of NCAR than the competitor Cray Research at that time. It was interesting to note that Steven Hammond from NCAR (National Centre of Atmospheric Research) was not allowed to display application benchmark results for the comparable Cray system at IEEE Supercomputing 1996. The reason was the controversy risen by NEC and Cray with regard to possible dumping by NEC. At NCAR the $ 35 million system should be installed. The dumping activities delayed the acquisition of the computer and did hurt NCAR badly.

On September 26, 1997, the ITC voted to impose hefty anti-dumping duties on imports of supercomputers from Japan. In a 3-0 vote, the commission found in favour of U.S. manufacturer Cray that the U.S. industry was being injured or threatened with injury by Japanese dumping of vector supercomputers. The Commerce Department set anti-dumping duties of 454 percent of their price against NEC, 173.08 percent against Fujitsu and 313.54 percent on all other Japanese firms.

The reasons of the Commerce Department to set such high duties were that NEC did not participate in the U.S. anti-dumping investigation and Fujitsu pulled out after a preliminary ruling.

Against this decision, NEC and its U.S. subsidiary HNSX Supercomputers appealed to the Court of International Trade against the September determination of ITC. NEC claimed that ITC excluded supercomputer sales by IBM, SGI and others. In November 1998, NEC filed a petition with the U.S. Supreme Court seeking review of a decision by the Federal Court. On December 22, 1998, the Court of International Trade overturned the finding by ITC. The judge found that ITC failed in showing a casual connection between the small number of Japanese Supercomputers sold in the U.S. and the deteriorating performance of Cray. ITC had to reconsider the dumping penalties within 90 days.

The final result was the U.S. Supreme Court rejected NEC and the Japan supercomputer appeal.

The implications have been noted four years later. The performance gap between the Crays and Japanese supercomputers became greater and greater. The same was true with the technology. The Cray T90s are based on ECL technology, while the Japanese vendors moved to CMOS. That made these machines cheaper, smaller and less power consuming. As late as in June 1998, Cray announced the SV1. The delivery was delayed and the application performance of the first systems was not as good as expected. Now there is a factor of nearly 7 in the peak performance between the "slow" SV1 and the high-end NEC's and Fujitsu's. The U.S. researchers had no access to these fast vector computers from Japan in their own country - because of the anti-dumping duties. End of 2000 Richard Rook from NASA published the findings of an ad-hoc working group on climate modelling. The results are that U.S. is falling behind other countries which use Japanese vector computers. The acquisition of such machines is the only chance to reduce the distance to the world.

Only two months later this became true by the cooperation of now Cray Inc. and NEC. The political anti-dumping decision costs the U.S. five years, as Japan never reacted in this way. In the last Top500 list from November 2000 shows that U.S. companies deliver 89% of those machines compared to only 11% of the Japanese firms. The Japanese dominate the vector computers as they sell all the 50 vector computers in that list.