Navigation Content
University of Wisconsin Madison College of Engineering
You are here:
  1. Home > 
  2. News > 
  3. News archive > 
  4. 1996 > 

UW Alums Team up at Cray to Build J90s

The parallel supercomputer on the tenth floor of the Engineering Research Building had ties to UW-Madison long before it was installed last May. No less than nine members of the team that designed this model, the J90, were once students in the College of Engineering.

This group of UW-Madison-educated engineers was formed early in 1992 when Cray Research, Inc., of the Twin Cities "recognized that a need existed for a smaller, lower-priced computer that would run the same code as the faster, more expensive machines," explained Ray Zastrow (BSEE '62), a senior logic engineer. "This was the birth of Entry Level Systems (ELS) within Cray."

Blue coats around Cray

Pictured in front of a Cray J90 computer system are UW-Madison College of Engineering graduates (from left) Tom Schultz, Ray Zastrow, Steve Berg, John Schomberg, John Pearson, Dan Chase and Steve Phillips. Not pictured are John Lien and Al Schiffleger. (large image)

Because there were not enough engineers at Cray to work on the new project, the company brought in a number of people from outside organizations. The J90 team eventually had about 30 members, divided into smaller groups for mechanical engineering, logic design, circuit design, software development and simulation work.

"It was a small, tight-knit team," said Steve Berg (BSEE '84), who helped design the memory interface on the central processing unit. "We had a loose structure but everybody knew their responsibilities on the project. There was a lot of personal one-to-one contact."

Since the first J90 was completed in December 1994, more than 250 units have been produced at Cray's Chippewa Falls, Wis., plant. This includes the Madison unit and another at UW-Stout.

According to David E. Foster, director of UW-Madison's Engine Research Center, the J90 has been "a tremendous asset to computational development." Previously, ERC researchers had to rely on using machines owned by other organizations. "We received some time gratis and we wrote proposals and established cooperative relationships wherever possible. But we were still always short on computer time.

"It's so much better to have our own machine."