THIS STORY HAS BEEN FORMATTED FOR EASY PRINTING
Innovation Economy

Through ups and downs, creative mill grinds on

By Scott Kirsner
January 4, 2009
  • Email|
  • Print|
  • Single Page|
  • |
Text size +

In January 1970, freshly graduated from college, Jud Leonard showed up for his first day at work at the old mill complex in Maynard. "I was so excited, I probably even wore a tie," he remembers.

His employer hired him to write software for one of the minicomputers it made. Leonard was among the first 9,000 employees of Digital Equipment Corp., which during Leonard's 16-year tenure grew to more than 100,000 employees, making it the second-largest computer company in the world, after IBM.

Now, a decade after Digital's sale to Compaq and slow dissolution, Leonard is again showing up every weekday for work at the old mill. As cofounder and chief architect at SiCortex Inc., he's designing a new generation of powerful computers that SiCortex markets as the world's most energy efficient.

If you want to understand the phenomenal regenerative ability of Massachusetts' innovation economy, head for the hulking brick mill in Maynard, close by the Assabet River. Carpets were made there in the 1840s, until an economic downturn, sparked by bank failures, killed that company. A new textile firm at the mill made blankets and uniforms for the military during the Civil War. That business failed in 1898, and was replaced by American Wool Co., which shut down its production facility in Maynard in 1950.

Starting in 1957, for more than three-and-a-half decades, the mill was Digital's world headquarters. In the 1990s, the career site Monster.com made the mill its base. And more recently, it has been home to dozens of start-ups, including the video game company 38 Studios, founded by Curt Schilling, and SiCortex, founded by Leonard and two fellow Digital alums, Matt Reilly and John Mucci.

In 2002, when they started SiCortex, Reilly had just left a job at Intel Corp., which had purchased Digital's microchip business, and Leonard had been consulting for Intel. Mucci had been an executive at Thinking Machines Corp., a Cambridge supercomputer company that lasted only a dozen years, filing for bankruptcy in 1994. He worked for a couple start-ups after that.

"Inside a large company, the amount of influence you can have is somewhat diluted," Reilly says of his time at Intel. "You can find yourself spending more time trying to build up your ability to influence things rather than building ideas." Despite the morose post-dot-com, post-9/11 business environment, Reilly and Leonard starting talking about ideas for their own company.

"We wanted to build a computer for scientific and technical customers - that was where our experience and our hearts lay," Leonard says. He adds that there had been very little innovation in high-performance computing "ever since Seymour Cray died" in 1996. (Cray, regarded as the father of supercomputing, founded Cray Research to build the world's fastest computers.)

Instead of relying on pricey supercomputers, by the late 1990s and early 2000s, computer users who needed lots of processing power simply took lots of Intel-based personal computers and lashed them together with Ethernet networking cables.

As the cost of PCs plummeted, it was an inexpensive but jury-rigged solution. Reilly and Leonard observed that one of the bottlenecks in these systems involved shuttling data from one PC to the others.

"We saw them as being hobbled because of the slow communication," Reilly says. They started to design their own chip, and integrated communications capability into it, rather than relegating it to a separate module. They developed their own communications protocol. And instead of connecting the chips with cables, they affixed copper wire onto the circuit board that held the chips, which gave them more communications bandwidth.

They also moved the chips closer together, which required them to start thinking about how to ensure that the chips would generate less heat than usual. "If you put a hundred or so PC processors in the space of one of our modules, they'd just melt down," Leonard says.

By considering how to make their chips work efficiently at a lower voltage, and simplifying the "thinking," or logic operations, that take place within the chip, they wound up with a computer that sipped power instead of gulping it. (The core team that designed SiCortex's chip came from Digital's Alpha chip development group, which had been based in Hudson.)

"By getting rid of the cables, we wanted to make the communications faster, improve the reliability, and make it easier to install," Reilly says. "The energy efficiency thing was, frankly, a surprise. It wasn't at the top of our agenda when we started."

But it gave the company a marketing edge. SiCortex didn't want to build the biggest, fastest, most expensive computers in the world. "Government labs may spend $30 million or $50 million on supercomputers from IBM or Sun or Cray, but those companies will often lose money on the deal," says Christopher Stone, who succeeded Mucci as chief executive in mid-2008.

For a desktop machine with 72 chips inside, SiCortex charges $20,000; the company's top-end machine, about the size of a small elevator car, has 5,832 chips inside and sells for about $2 million. (It also has a nifty gull-wing door and flashing blue LED lights, both of which remind me of the DeLorean time machine from "Back to the Future.") And with electricity representing a major cost of keeping the machines running, Stone boasts that SiCortex computers use 60 percent less juice than anything else out there.

How well the pitch will work remains to be seen, though Stone claims 2008 was a successful year. "We shipped 60 computers, and this was the first full year they were available," he says.

Early customers include schools like MIT, Boston University, and Princeton, along with NASA and General Electric. The company has raised about $60 million so far from investors, including the local venture capital firms Flagship Ventures, Polaris Venture Partners, and Prism VentureWorks. But Stone says the company will need to raise more funding this year to develop a second-generation computer that will be able to run a wider range of software applications.

"Do I worry about raising money in this environment? Sure," Stone says. "But we're generating revenue, and the customers want to buy more."

With a bit of pride, Stone says SiCortex is one of the first companies in Massachusetts to design a new computer from the ground up since the 1980s heyday of Digital, Thinking Machines, and Apollo Computer.

With 79 employees, SiCortex is starting to outgrow its current space in the mill - but Stone says there's room for expansion there.

When I visited last month, Reilly shared a bit of Digital lore with me - the rumor that when the company built a computer that didn't perform as expected, it was heaved out a window into the mill pond. He says SiCortex hasn't yet had to do the same sort of quality-control-by-defenestration.

"Every time I turn a corner, I see something I remember from 30 years ago," Leonard says. "It's strange and kind of wonderful. In some ways, it feels like that same excitement of starting my first job - the same excitement comes back."

Scott Kirsner can be reached at kirsner@pobox.com.

  • Email
  • Email
  • Print
  • Print
  • Single page
  • Single page
  • Reprints
  • Reprints
  • Share
  • Share
  • Comment
  • Comment
 
  • Share on DiggShare on Digg
  • Tag with Del.icio.us Save this article
  • powered by Del.icio.us
Your Name Your e-mail address (for return address purposes) E-mail address of recipients (separate multiple addresses with commas) Name and both e-mail fields are required.
Message (optional)
Disclaimer: Boston.com does not share this information or keep it permanently, as it is for the sole purpose of sending this one time e-mail.