SPARC at 25by Diana Reichardt
Computer History Museum event delves into the question: Why SPARC?
In 1987, a small startup called Sun Microsystems developed its own microprocessor, called SPARC, and introduced the Sun-4, the first computer based on the new chip. On November 1, 2012, many early SPARC team members, along with Oracle President Mark Hurd and Executive Vice President of Systems at Oracle John Fowler, convened at the Computer History Museum in Mountain View, California, for SPARC at 25: Past, Present, and Future. This event provided a look back at the history of SPARC and the early days of Sun. Stories from the participants, including all of the company’s founders, illustrated the complex nature of systems design and the challenges of launching a high-technology company in a fiercely competitive industry.
While the event was filled with details on the challenges the team faced in the late 1980s, one topic prompted a discussion that surprised even the veterans of Silicon Valley: Why SPARC?
Panelist Vinod Kholsa, a Sun cofounder and now a venture capitalist, said that the only way Sun could compete at the beginning was to use open source systems that allowed the company to leverage everybody else’s work. “We had never done a chip before,” said Kholsa, and he related that there wasn’t a lot of interest in doing SPARC initially, but the team realized they needed a way to differentiate and beat the competition.
Panelist Andy Bechtolsheim, cofounder and chief system architect at Sun during the development of the SPARCstation series of workstations, said the team knew the risk was enormous, but he added that “if you don’t take some of those bets, you’re never going to be great. It could’ve broken the company, but the will to go and get it done was there.”
Sun cofounder Scott McNealy, who appeared at the event in a recorded video, said that when Sun launched SPARC everyone said, “What are you trying to do? How can you compete against Motorola, IBM, Intel, and all of the others in the microprocessor space?” The answer was simple, he said. Nobody was delivering the kind of floating-point performance and technical computing power necessary for engineering workstations. So Sun decided to design its own RISC architecture and named it SPARC for scalable processor architecture.
McNealy said the team knew they had a hit when they got the first chip back. “It was pretty clear from a price-performance standpoint that we had a huge winner,” he said. “We had a magical cocktail that exploded in the technical marketplace.”
Panelist Bill Joy, a Sun cofounder and a key member of the architecture and software teams, said that SPARC architecture was needed because he didn’t see support for symbolic computing in an instruction set. “I wanted a hardware platform that would support the future direction of software,” he said.
One panelist continues to work on the design and performance of SPARC processors today: Rick Hetherington, vice president of hardware development at Oracle. He joined Sun in 1996 as coarchitect of Sun’s Millennium processor project.
“The Millennium project was the early part of threading superscalar machines,” he said. When asked what the future of SPARC holds, Hetherington replied, “More cores, more threads—we’ll continue to turn up the clock.”
Nathan Brookwood, of semiconductor market research firm Insight 64, attended the 25th anniversary event as well as the initial launch of SPARC in 1987. “When Scott McNealy stood in front of a press conference 25 years ago and said SPARC would revolutionize the industry, few in the room, including me, took him seriously,” he said in an EE Times story reporting on the anniversary.1 “But SPARC has now proven that it has some staying power.”ADD FOOTNOTE HERE
Diana Reichardt is a senior writer at Oracle.