Tilera, Lonely Survivor in Multi-Core Chip Business, Gathers $45 Million for Push Into Cloud Computing

(Page 2 of 2)

the individual cores run at slower clock speeds.) “With half of the expense of running a data center being power, this is very attractive,” says Bailey.

Tilera has 85 employees spread across offices in the U.S., China, Japan, and Ireland. Most of the company’s chip designers are in Westborough, while the sales, marketing, and executive staffs are in San Jose. The company expects to become profitable in the second half of 2011, but the new funding will allow it to scale up sooner than that. “One of the key purposes of this round is to expand the team,” especially in sales and marketing, Bailey says. “We’re in a number of markets and we’re gaining a lot of traction, but we need more feet on the street.”

Tilera intends to put part of the money toward its next-generation products: a 100-core chip coming out later this year, followed by 225-core chip expected in 2013. Bailey says it’s relatively easy for the company to add more cores to a chip or “tile” thanks to their mesh-networked architecture, in which each core is connected to four neighbors via multiple “buses” or wires dedicated to passing cache data, system messages, and other information. Agarwal developed the technology at MIT with DARPA and NSF funding, and Tilera has an exclusive commercial license.

“A 100-core [chip] has more components on it than a 64-core, but it’s not anything different,” says Bailey. “We think it can scale up to 1,000 cores before you start to see big issues.” At that size, chips would need so much memory and so many wires for input/output operations that they’d become unmanageably large, and it would take too many clock cycles for data to travel from one side of the chip to the other. “But we don’t see those things as issues anytime soon,” Bailey says.

A number of specialty chipmakers such as BrightScale, Phiar, and WiQuest have shut down over the last few years, beset by the recession, the expense of creating new chip designs, and the difficulty of winning adoption by builders of mainstream consumer or enterprise hardware. Ambric, a Beaverton, OR, startup founded around the same time as Tilera and once seen as a potential competitor, went belly-up in 2008 after raising at least $33 million for its multi-core digital video processing chips. And Stream Processors, another competitor based in Sunnyvale, CA, closed in 2009 after five years in business; it had raised more than $26 million in venture funding for its own parallel processing architecture.

But after a tough 2008-2009, the chipmaking business is rebounding: global semiconductor sales surged to $304 billion in 2010, up from $230 billion the year before, according to market research firm iSuppli. That rising tide may be lifting all boats, including specialty chipmakers. “It’s nice to have some good news in the semiconductor industry, with so many companies shutting down in the last couple of years,” Bailey says of Tilera’s fourth funding round. “Having all these investors be excited and making this round possible is a very good sign for the industry.”

Single PageCurrently on Page: 1 2 previous page

Wade Roush is a freelance science and technology journalist and the producer and host of the podcast Soonish. Follow @soonishpodcast

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

Comments are closed.