Directly from an employee in the Test and Engineering departmen𝄹ts of Cray Research and lateღr Cray Computer

THE ONLY SIGNIFICANT CR🔯AY MANUSCRIPT TO COME TO AUCTION

Cray Computer Corporation employees would start a new logbook for each new module build that was received for testing and Seymour ✃Cray follows that protocol in the present logbook. The logbook would remain in the test department and be retrieved each time it returned from repair. 

The logbook opens with a dated entry: “Sept 5, 1989 SR Cray 10AM / New Module build, Stacks… / clock distribution looks ok in stacks… / (TA) package pulse looks ok. Wide clock and voltage margins. / Using new Macintosh interface with transmit board….“  

The logbook continues with Cray’s meticulous and orderly notes that span five months with gaps when the module💯 was out for repair. 

Cray seems to have learned all he could with this module by his final entry, January 30, 1990: “Not clear what more can be done with vector registers at this point. All of [?]6 parts are overwriting previous elements in this wafer lot. Work has begun on VR7 Assign to put capacitors back in with readout delay of VR6. / Assume this module proceeds with scaler register checkout until another wafer lot of VR package is available.” 

Cray ap𒁃parently handed off the module to a technician at this point as the remaining 41pp are in another hand♏. 

Cray had achieved legendary status with his designs at Control Data Corporation (CDC) and especially his first two Cray supercomputers, the Cray 1 and 2. The Cray 3, which had already begun designing before the Cray 2 was complete, used some of the innovations from earlier designs like vector proc꧒essing, 3D integrated circuit packaging and immersion in the inert liquid Flourinert for cooling, but Cray was one to take large technological leaps. He aimed for a 12x increase in speed over the Cray 2. He would achieve this with a number of innovations including greatly decreasing the pathways by stacking 9 printed circuit boards containing 69 electrical layers and the use of gallium arsenide chips. 

The present logbook was completed at a time when the Cray 3 was far behind schedule, and Seymour Cray was doing everything he could to speed the glacial process including working long hours – and even turning down the National Medal of Technology as he would have had to have gone to the White House to accept it. There was also the fact that Cray wanted to keep his hands in the nuts and bolts of the operation: “At sixty-five, Cray was still the lead designer on the Cray-3 project. He wrote the Boolean equations, designed the board interconnects, and defined the so-called wiring masks. It wasn’t unusual to see Cray sitting at a bench in one of the labs, doing the microprobing of the tiny circuits on the modules, testing for signal errors. In the beginning, technicians and assemblers were stunned to see this wealthy genius hand-wiring the modules. Cray did the most monotonous and tedious tasks, partly because he felt it made him understand the machine better, and partly because he felt that no engineer should be above it” (Mur🗹ray𓄧 p. 204). 

One oꦓf the issues was that GigaBit Logic was the only manufacturer of gallium arsenide chips and each time a new design was sent to them it took three to four months for it to be completed. The difficulties had already led to Cray and Cray Research parting ways in mid-1989. Cray would continue the Cray 3 project with a large investment൲ from Cray Research under his new company, Cray Computer Corporation. The setbacks led to the cancellation of the launch order by Lawrence Livermore National Laboratory in 1991. Not a single unit would be sold although one was loaned to the National Center for Atmospheric Research (NCAR) when it was finally completed in 1993. Cray went on to design further systems before he died in a 1996 automobile accident, but none were brought to market. 

REFERENCES:

Murray, Charles J. The Supermen. 🅘New York: John Wiley & Sons, [19🐈97].