OK, this deal wins the award for clunkiest company names—but it’s an important exit for a Boston-area startup in a tough market. CHiL Semiconductor, based in Tewksbury, MA, confirmed today it is being acquired by International Rectifier (NYSE: IRF), based in the Los Angeles area, for $75 million in cash. The deal is expected to close before the end of March.

CHiL Semiconductor, which was founded in 2006, designs computer chips for power management in servers, PCs, and high-performance computing systems. The technology’s main applications are in graphics, gaming, and other high-volume industries.

In a statement, CHiL’s CEO, Ram Sudireddy, said International Rectifier is “an ideal partner for our customers and employees.” He added that the merger could help “create a higher performance and lower cost” technology that would enable end users to create more energy-efficient products.

The sale seems like a decent exit for CHiL’s investors, which include Flybridge Capital Partners, Highland Capital Partners, IdSoft Capital, and Panorama Capital. CHiL has raised a little more than $30 million since its founding. The most recent financing was a $16.7 million Series C round, led by Panorama, in 2008.

International Rectifier was founded in 1947. The venerable power-management firm has manufacturing facilities in Leominster, MA, among other places.

Comments | Reprints | Share:          

Chelmsford, MA-based Voltaire (NASDAQ: VOLT), a developer of servers, switches, and software for high-performance and cloud computing environments, announced today that it has agreed to be acquired by Mellanox Technologies for $8.75 per share, for a total equity value of $218 million. The deal is expected to close first quarter of 2011 and will help Mellanox (NASDAQ: MLNX) strengthen its position in the end-to-end connectivity for data centers and servers space, according to the announcement. The company said it will bring the Voltaire employees under its management team, retain existing product lines, and run both companies from its offices located in Israel, the U.S., and around the world.

Comments | Reprints | Share:          

Seattle-based Amazon Web Services announced today it has introduced a new cloud-computing service based on graphics processing units (GPUs). Portland, OR-based Elemental Technologies is the first company to offer cloud-based services for video transcoding on top of Amazon Web Services.

Elemental is going after broadcast and online video customers in its push to become the go-to supplier of technologies that help people watch high-quality video on any device, anytime.

The target customers of Amazon’s new service, more broadly, are developers and businesses that need high-performance computing on demand. GPUs are powerful graphics chips that provide unique parallel-processing capabilities. Amazon Web Services is using chips made by Nvidia.

GigaOm reported more details and context around Elemental and Amazon last month, well ahead of any announcement.

Comments | Reprints | Share:          

Thomas Thurston is a startup predictor. Tell him about your company, and he’ll tell you whether it will survive or fail.

No, he’s not an investor, or a psychic. By day, Thurston is a mild-mannered researcher and consultant whose training is in law and business. He’s the founder of Portland, OR-based Growth Science International, a research firm that works with entrepreneurs, investors, and corporations on their business strategy. By night, though, he’s testing every possible angle of a theory that could change the way a lot of people think about startup strategy.

Here’s the upshot of Thurston’s recent research, and why it’s important. Pretty much every startup you’ll ever meet will say it is better than its competitors. However you want to measure it—speed, technology, revenue model, whatever—a young company will say it outperforms others in its class. What’s more, it’s smaller and nimbler than the big companies, so it will be able to innovate faster and stay ahead of the curve.

Just one problem: That’s exactly why it will fail.

What a startup should do instead—to give itself the best chance of surviving—is enter the market at the low end of performance, Thurston says. That is, offer a product that’s not necessarily as good as its competitors, but is cheaper and more accessible. “Lower cost, lower performance, and gets better over time,” is how Thurston puts it.

If this sounds familiar, you’ve probably read Clayton Christensen’s books on business innovation. Christensen, a Harvard Business School professor, is the author of The Innovator’s Dilemma, The Innovator’s Prescription, and Disrupting Class, and he is coming to Seattle on May 17 to give the keynote at the Technology Alliance’s annual State of Technology Luncheon. The connection to Thurston is that he and Christensen have collaborated on testing predictions about startups and other companies.

In 2005, Thurston was working at Intel Capital when he got interested in whether a mathematical model could predict startup success or failure better than chance. He plowed through obscure academic papers and popular books, tried different things, and settled on building a sophisticated model based on Christensen’s principles of “disruptive innovation” (more on this definition shortly). Thurston got a hold of 48 business plans from within Intel—new businesses that had corporate funding—and checked how they did (survive or fail) against what Christensen’s model would predict. To his surprise, the model made accurate predictions more than 85 percent of the time, and the results were highly statistically significant.

Thurston decided to take a year off from his job in 2007 to continue the research with Christensen in Boston, co-sponsored by Intel and Harvard. They expanded their analysis to include all new businesses Intel has supported (roughly 100), as well as hundreds of outside companies across different industries and geographies. The result was the same: 85 percent accuracy.

Skeptics would say the model was tested by its own proponents, so it’s not surprising they would find it accurate. But Thurston maintains he is an independent researcher; he would happily switch to another model if it worked better, he says. He has since returned to Portland and continued the work at Growth Science, where doing the modeling is part of his consulting gig. He says he’s been getting lots of interest from companies and venture capitalists seeking advice.

So here’s how the predictions work, in a nutshell. First, a company is classified according to whether its market strategy is “sustaining” or “disruptive.” Sustaining means it is positioned as … Next Page »

Comments (15) | Reprints | Share:          

Seattle-based Cray, the supercomputing company, announced today it has won three high-performance computing awards from the U.S. Department of Defense, totaling more than $45 million. Cray will provide three next-generation supercomputing systems to the U.S. Air Force Research Laboratory in Ohio, the Arctic Region Supercomputing Center in Alaska, and the U.S. Army Engineer Research and Development Center in Mississippi. Cray (NASDAQ: CRAY) says its technology will be used to support research and development for new materials, fuels, and armor and weapons systems, as well as to assist in long-term weather predictions. The supercomputers are expected to be delivered to the U.S. defense centers in the second half of 2010.

Comments | Reprints | Share:          

The demise of San Diego’s Verari Systems and its resurrection as Verari Technologies happened so fast you might not have realized what happened unless you were paying close attention over the holidays.

The longtime local company, which specializes in making high-performance servers, server racks, and energy-efficient data storage centers, laid off 223 employees and ceased operations just two weeks before Christmas. It was hard to make out what was happening at the time, chiefly because Verari denied media reports that it had shut down—saying instead that its doors were open and it was merely “restructuring” its business.

But a few weeks later, a notice on Verari’s website said the high-performance computer company was in fact soliciting buyers in a public auction. All Verari assets were later sold through a liquidation known as an “assignment for the benefit of creditors.” Such liquidations, sometimes known as “friendly foreclosures,” are not that unusual—especially these days, according to Bruce Bennett, a Los Angeles bankruptcy lawyer. Bennett told me that in these scenarios, a senior secured creditor, usually a bank that provided a business loan, forces the issue, but ultimately with the company’s cooperation. “It’s just an inherently simpler and less expensive way of liquidating the assets instead of going through bankruptcy,” Bennett says. In most cases, no new company emerges from such liquidations.

But Bennett says it’s not that unusual for a key figure from a distressed company to step back in and acquire all the assets—which is what happened at Verari. Dave Driggers, who founded the San Diego company in 1991 and was working as Verari’s chief technology officer, led an investment group that bought all of the old Verari’s assets. Driggers also stepped in as CEO to reboot the San Diego computer company, which is now known as Verari Technologies. Less than 40 percent of the old Verari’s 223 employees are expected to get their jobs back.

Dave Driggers

Verari reopened for business last week, capping a lightning-fast salvage that took less than six weeks to carry out. But it tortures common sense to describe what happened as a restructuring.

In a recent interview, Driggers told me that all of Verari’s previous investors were wiped out—after raising $59 million in three rounds—and that they were not part of the group that bought Verari’s assets. “You never heard me announce a restructuring,” Driggers said. “I chose not to say anything during the process—from the day we shut down until after the acquisition.”

During our conversation, Driggers explained why he’s optimistic about the future of Verari Technologies—which he says is moving forward with the same technology and a better business model. He also explained why the old Verari failed.

With a mixture of exasperation and pride, Driggers said that Verari “had incurred a huge amount of debt” while its technology and products were … Next Page »

Comments | Reprints | Share:          

A new IBM supercomputer opened at Brown University’s Center for Computation and Visualization in Providence, RI, today is 50 times faster than Brown’s next best machine and is the most powerful computer in Rhode Island, according to an announcement from IBM (NYSE: IBM). Researchers at Brown and other institutions intend to use the 1,440-processor machine to model subjects such as the genomes of ocean-going microbes, the mechanics of human and animal movement, and the topography of other planets. Brown ordered the multimillion-dollar supercomputer in June; its exact cost hasn’t been disclosed, but IBM and Brown are calling it “a shared investment.”

Comments | Reprints | Share:          

Well, we can’t all be Amazon. While the Seattle-based e-commerce giant (NASDAQ: AMZN) raked in a $199 million profit for the third quarter of 2009—a 68 percent increase in net income over the same period last year—Microsoft (NASDAQ: MSFT) posted an 18 percent decline in its profits (still $3.57 billion, better than analysts expected).

But beyond these giants of the global tech scene, Seattle has some mid-market public tech companies that we’ve been paying closer attention to lately. That’s because they provide a much more complete picture of what’s going on in the public markets, as well as the mood across different industries like digital media, data storage, and high-performance computing.

Of these local bellwethers, two companies announced modest quarterly profits this week, and two others posted losses but are on the long-term comeback trail. It’s clearly still tough times out there, but here are the highlights:

—RealNetworks (NASDAQ: RNWK), the Seattle digital media and entertainment company, managed to post a surprising profit of $1.5 million for the third quarter of 2009, its first profitable quarter since the first three months of 2008. That’s despite posting quarterly revenue of $140.3 million, a decrease of 8 percent from $152 million in the same period last year (when the company posted a net loss of $4.5 million). RealNetworks reduced its operating costs and formed partnerships with Facebook and Apple over the past few months.

—Cray (NASDAQ: CRAY), the Seattle-based supercomputing company, reported a net loss of $2.1 million for the third quarter. But its revenue was $58.6 million, a 7 percent increase over the same period in 2008. In the second quarter of this year, Cray posted a surprise profit of $3.4 million on the strength of large government contracts and a broader customer base.

—InfoSpace (NASDAQ: INSP), the meta-search company based in Bellevue, WA, posted a profit for the second straight quarter (following three consecutive quarterly losses). Its net income for the third quarter was $1.8 million, based on revenue of $54.4 million, an increase of 38 percent over its revenue from the same period a year ago.

—Isilon Systems (NASDAQ: ISLN), the Seattle-based data storage firm, reported a net loss of $4.9 million for the quarter. The company’s quarterly revenue was $30.5 million, up 1 percent over the same period a year ago, but its net loss increased from $3.7 million in the previous quarter this year. I wrote about nine-year-old Isilon’s efforts to bounce back from some tough times in a profile last week.

On October’s last trading day, the stock market plunged. As Scott E. Marcouiller, a senior equity market strategist at Wells Fargo Advisors in St. Louis, told the New York Times today, “The market is focusing on the glass is half empty…We just needed to let some of the air out of the balloon.”

Comments | Reprints | Share:          

The San Diego Supercomputer Center at U.C. San Diego today named UCSD astrophysicist Michael L. Norman as interim director. Norman replaces Fran Berman, a specialist in grid computing (and San Diego Xconomist), who has served as the center’s director since 2001. Berman plans to join the Rensselaer Polytechnic Institute as vice president of research, effective Aug. 1. Norman, a UCSD distinguished professor of physics, was appointed as the center’s chief scientific officer a year ago. The university says it plans to name a permanent director at a later date.

Comments | Reprints | Share:          

Civic leaders, economic visionaries, and passionate residents often claim that their city is the next to rebound. I say with confidence that my city is next in line for metamorphosis. Some find their new meaning through the arts and other organic movements, but in most recent accounts, it seems, a few deserving cities are staged for a comeback with the help of high-tech industry. Holyoke, Massachusetts, is one of these cities.

Last week, it was announced that the Massachusetts Institute of Technology, University of Massachusetts, and Boston University would be collaborating with Fortune 500 technology giants Cisco Systems and EMC and the Commonwealth of Massachusetts to build a high-performance computing center (HPCC) in the city of Holyoke.

I got involved in the early stages of this effort back in October of 2008, when I was invited to a Technology Roundtable at MIT. It was surreal finding myself, a young web entrepreneur, talking with and having breakfast at the same table as Governor Patrick, MIT president Dr. Susan Hockfield, Cisco’s John Chambers, and various other leaders and innovators in the world of business and technology. Major topics included the future of the tech industry in Massachusetts, its impact on the state economy, and support of entrepreneurship. When the conversation moved in my direction, I was encouraged to share my thoughts and suggest various underutilized opportunities I saw as a young entrepreneur living and working in the western part of the state. I made what I hoped were persuasive statements for my city, advocating that it would be the premier location for sustainable, cost-effective, large-scale data centers. By the look on everyone’s faces, it was clear that Holyoke had been on their radar for such a project.

Western Massachusetts has many valuable resources that would make it an appealing and economically advantageous destination for the tech industry. But what firmly stands out in this region is the small city of Holyoke and its abundance of power—specifically, cheap, renewable hydro-electric power. The city receives approximately three-quarters of its energy from its hydro-electric sources, which include an expansive dam and turbines located along canals powered by water diverted from the Connecticut River. The municipality-owned utility company also announced recently that it would be acquiring hundreds of acres of land on a nearby mountain range to develop a high-capacity wind farm. In regards to technology, Holyoke already has an extensive high-speed fiber optic network throughout its downtown and is located along the Mass Information Turnpike backbone.

It is likely that Holyoke’s new computing and data center will be modeled after Google’s “Project 02″ in The Dalles, OR, and its data-center in Baudour, Belgium. Water from the river or industrial canals will be used to cool the servers and ensure that the hardware runs efficiently.

When it was founded in the mid-19th century, Holyoke was one of the first cities planned for industry—and it promoted its plentiful supply of inexpensive energy and labor to lure manufacturing businesses and investors. It became the center of the global paper industry and spurred tremendous amounts of wealth, laying claim to being the richest city per-capita in the United States by the 1920s. At the city’s prime, Holyoke’s cultural offerings were compared to those of New York City and Paris. After the Great Depression and general de-industrialization that struck the Northeast, Holyoke slowly began to lose its prosperity, businesses, jobs, population, and reputation. For as long as I can remember in my 21 years of existence, the city has been in rough shape. People often blame the city’s reputation as what’s repelling business—but real innovators, entrepreneurs, and “urbanists” see past that image and recognize this diamond in the rough. The raw, undisputed assets stare you straight in the face—and, finally, the right people are staring back. This new computing collaboration will be an anchor, attracting entrepreneurs and young minds eager for involvement in something of great magnitude.

As this high-profile project moves forward, I’d expect many eyes from the technology sector to be fixated on Holyoke. There is still over two-million square feet of vacant building space, consisting mostly of red-brick factory structures ready to be reclaimed by the high-tech community at bargain prices. Is there a chance that a company like Google might look at Holyoke for its next data-center in the Northeast? I believe it’s very likely.

I’m optimistic about what the future holds for the city of Holyoke. High tech offers the much-needed flow of capital, job creation, and stimulus that post-industrial cities like Holyoke desperately thirst for. With the escalating focus on green industry, sustainability, and long-term cost factors, Holyoke is a premier location for the Northeast IT sector. The beloved Queen of Industrial Cities as it was called at turn of the century, fell off her throne many decades ago, but is now back with a new purpose. Thanks to the forward-thinking powers at hand, Holyoke will be given a new identity as a coveted destination for high tech.

Comments (7) | Reprints | Share:          

Updated with additional detail June 9, 8:10 pm; further update and a correction added June 10, 7:15 am—see below

A university, industry, and Massachusetts government coalition is set to announce a major initiative in high performance computing later this week, Xconomy has learned. Official sources are guarded, and specifics were hard to come by tonight. But a new center for high performance computing is apparently being planned for Holyoke, MA. A public event is planned for Thursday, with representatives of the state of Massachusetts, MIT, EMC, and a variety of others due to be on hand, according to our sources.

This paragraph added, June 9, 8:10 pm: The presidents of the University of Massachusetts and MIT, as well as leading executives of Cisco, are also involved with the effort, we have learned. And a centerpiece of the announcement is the creation of a data center that will bring jobs to Holyoke, an economically depressed region, our sources say.

Kofi Jones, director of Communications and External Affairs for the state Executive Office of Housing and Economic Development, confirmed that an announcement on high performance computing was scheduled for Thursday at Mt. Holyoke, but did not provide further details by the time of this post. We could not determine whether federal funding is part of the arrangement.

The new computing center is also intended to blaze trails in so-called “green” computing, according to our sources. This usually refers to getting more performance per kilowatt of electricity from computing systems. High performance computing systems are notorious energy hogs.

This section added, June 9, 8:10 pm: “It’s a big data center. What they’re trying to do is bring to Massachusetts companies like Google and Microsoft,” said a source, whom we promised not to identify. The source referenced the large data center operated by Google along the Columbia River in Oregon. “They’re trying to replicate this, I think, in Holyoke.” Holyoke is built on the banks of the Connecticut River.

A vigorous job training program is also part of the effort, our source said.

Update, June 10: An article this morning in the Boston Globe, which had apparently arranged with state officials to have an exclusive on this story before our article came out, confirms the basics of our report, but adds more details. It says the effort is expected to cost roughly $100 million, that it will seek federal stimulus funding, and that it will be managed by a collaborative led by MIT and the University of Massachusetts, with participation from EMC and Cisco. “At this stage, the partners have only agreed to participate in an intensive 120-day planning project that’s intended to work out details such as siting, organization, and funding,” the Globe says. “But the backers say the computer center will create an important resource for the state’s high-tech industry and academic institutions.”

Correction: A reference to Mt. Holyoke College in our original post was deleted as it became clear the center is being planned for the city of Holyoke and not the college as one source had seemed to indicate.

Comments (3) | Reprints | Share:          

SiCortex, a six-year-old startup building energy-efficient supercomputers in Maynard, MA, has shut its doors. The company ran out of working capital and was unable to raise more from its venture investors, according to a report that surfaced yesterday in HPC Wire, a trade publication in the high-performance computing industry.

Xconomy obtained confirmation of the shutdown news from John Mucci, the founding CEO of SiCortex, who was replaced 10 months ago by current CEO Christopher Stone. Stone himself was not immediately available for comment.

“It is a sad day for all… less competition, unemployed seventy some workers…” Mucci said in an e-mail.

SiCortex had raised $42 million in venture capital, including a $21 million Series A round in 2004 contributed by lead investor Polaris Venture Partners and syndicate members Flagship Ventures, JK&B Capital, and Prism Venture Partners. All of those investors returned for the company’s $21 million Series B round in 2006, with the addition of new lead investor Chevron Technology Ventures. Bob Metcalfe, a Polaris partner frequently quoted about SiCortex in the past, said this morning he had no comment about the shutdown.

According to the HPC Wire report, which was based on information from an anonymous source close to the company, most of the company’s employees have been let go, and a sale of the company’s assets is underway.

SiCortex had been in the process of raising additional working capital; according to documents filed with the Securities and Exchange Commission on May 13, it planned to sell $1.1 million in subordinated convertible promissory notes. It’s not known whether that debt fundraising was completed.

SiCortex had developed a line of large, multiprocessor, parallel computers that saved electricity by using somewhat slower processors, but made up the speed difference using an advanced internal communications network for shuttling data between processors. The company targeted customers doing technical computing work in areas such as geological exploration, climate simulation, and intelligence. Sales were apparently brisk; the company had announced record growth in the first quarter of 2009. The Royal Military College of Canada was one recently announced customer.

SiCortex had also been emerging as a leader in the field of “green” high-performance computing. Last November it introduced the “Green Computing Index,” a list that ranked the world’s largest supercomputers not according to their sheer speed but according to their power efficiency (expressed in billions of floating point operations per second per kilowatt consumed).

We expect to obtain further details about the shutdown later today.

Comments (1) | Reprints | Share:          

There’s no such thing as a free flop.

A “flop,” or floating-point operation, is a calculation involving a decimal number; engineers often measure the performance of computers in terms of the number of “gigaflop/s” (billions of floating-point operations per second) they can sustain. In the supercomputing community, it’s a longstanding ritual to compare various machines’ maximum gigaflop/s as they tackle standardized math problems called benchmarks. The world’s fastest computer—a 122,400-processor IBM machine at the Los Alamos National Laboratory called Roadrunner—can run at just over a million gigaflop/s, or 1.026 petaflop/s, according to Top500, the most prominent list of top supercomputing sites.

But as a general rule, the faster a computer runs, the more power it consumes—and the more waste heat it generates, and the more additional power is needed to run cooling systems. If current trends continue, according to McKinsey & Company, then by 2020 the electric plants needed to power the world’s data centers will be churning out more greenhouse gases than the entire airline industry. The problem is getting so serious that some organizations are having to scale back plans to upgrade their data centers with faster machines—not because they can’t afford them, but because local utilities can’t supply any more electricity.

Maynard, MA-based SiCortex serves the high-performance computing market, so it’s naturally obsessed with the gigaflops game—but with a twist. The six-year-old startup builds massively parallel computers with thousands of processors. The processors themselves aren’t very fast. They run at around 700 Megahertz, slower than the chips inside most desktop and laptop PCs, which saves a lot of electricity. But they’re wired together in a way that makes SiCortex’s computers extremely zippy nonetheless. And today SiCortex is proposing an overhaul in the way the performance of high-end computers is measured and ranked, one that would take a machine’s power consumption into account and reward machines that use it sparingly.

The company calls its new measuring system the Green Computing Performance Index, and it’s urging managers of government and academic supercomputing centers and corporate data centers to use it to evaluate the full benefits and costs of owning high-performance computing systems from companies like Cray, SGI, Hewlett-Packard, IBM, and of course, SiCortex itself.

When evaluating the full cost of owning a high-performance computer, SiCortex argues, organizations should divide its performance in gigaflop/s by its power consumption in kilowatts. When you do that, a number of machines that are nominally faster than SiCortex’s machines—such Cray Inc.’s XT3 and XT4, IBM’s Blue Gene, and SGI’s Altix 8200EX—come out looking like power hogs. When running the standard Linpack benchmark, for example, an 1,100-processr Cray XT3 machine at the Swiss National Supercomputing Centre gets just 17 gigaflop/s to the kilowatt, while a 1,458-processor SiCortex machine gets a comparatively huge 253 gigaflop/s to the kilowatt.

“If you look at the high-performance computing benchmarks that Top500 produces, it’s great stuff, but it doesn’t give you a measure of the actual energy efficiency of the computers themselves,” says Christopher Stone, SiCortex’s CEO. “Everyone in the high-performance computing business is running around talking about being green or wanting to be green, so we thought why not … Next Page »

Comments (2) | Reprints | Share:          

The new five-story building that doubles the size of the San Diego Supercomputer Center is more than just concrete, steel, and glass. As Fran Berman, SDSC director (and an Xconomist) said yesterday, the new building also represents the next generation in supercomputing—and she wants to ensure that San Diego’s emerging companies feel like they’re part of the family.

Berman ostensibly was hosting a dedication ceremony for the SDSC’s $44 million expansion. But the event included breakfast for more than 100 local technology executives, venture capital partners, and others to explain how supercomputing is crunching massive amounts of data to analyze Medicare fraud, cancer cells, earthquakes, and climate change.

Such research may not necessarily be useful to the venture community, but biotech and information technology startups increasingly need such capabilities, especially in digital storage and data mining, and to analyze complex models and simulations.

Today “SDSC sees its role as a partner and pathfinder with the research and education community to harness the power of cyberinfrastructure to adderess science and society’s most critical problems,” Berman said.

The supercomputer center, which was founded on the UCSD campus 23 years ago, now boasts that it has more archival storage capacity than any other academic institution in the world, about 25 petabytes (25 thousand trillion bytes). That’s about 1,000 times the digital plain-text equivalent of the printed collection in the Library of Congress.

The new building also houses some of the first optical networking equipment to be connected to the TeraGrid, a next-generation Internet for scientific research funded by the National Science Foundation.

“I have a number of models where we could work with industry,” said Ron Hawkins, SDSC’s director of industry relations.

The center could conduct sponsored research to address specific questions a company may have, Hawkins said. SDSC researchers also can do “high-end consulting” on corporate projects, or the center can simply offer access to its computational resources.

As one industry consultant noted, however, the programming required to solve complex problems on high-performance computers can be extremely expensive.

“Parallel processing is a specialized field, so, yes, it requires specialized expertise,” Hawkins acknowledged. “But the computing world and the IT world is going to parallel processing, so it’s only appropriate for us to work with industry to bring them along.”

Comments | Reprints | Share:          

The Assabet River these days rushes through Maynard, MA, without lending any of its liquid muscle to local industry. But for more than a century, the river supplied power to the Assabet Woolen Mill, a vast brick complex that, in its heyday, was the largest source of wool for U.S. military uniforms. I went to the mill two weeks ago to visit computer maker SiCortex, which is just one of numerous high-tech startups, including Monster.com and 38 Studios, that have taken over the complex, now known Clock Tower Place. And when I saw how swiftly the Assabet flows past the old mill buildings, I was reminded that for some companies—including, increasingly, computing companies—rivers are still a prime source of power. Google, for example, spends so much money on electricity that the search giant decided to build its newest data centers near hydroelectric dams in Washington state, where electricity is cheaper.

As it turns out, SiCortex’s whole mission is to help organizations do lots of computing without having to worry so much about energy costs. The company makes massively parallel computers that contain thousands of individual processors, wired together in a way that lets them exchange data very quickly—so quickly that the processors themselves don’t have to be very fast in order for the machine as a whole to carry out trillions of operations per second. And because the processors in SiCortex’s machines run at a relatively pokey 700 Megahertz, they don’t consume nearly as much power (or give off as much waste heat) as the multi-Gigahertz processors hawked by the Intels of the world.

If you take power and cooling expenses into account, according to SiCortex, its machines are only one-third as costly to own and operate as equally fast Intel-based clusters. In fact, a SiCortex machine uses so little electricity that it can be powered by a small team of cyclists. The company organized just such a stunt at MIT last December, when 10 members of the MIT cyclocross team hooked stationary bikes up to generators and pumped out enough juice to run a fusion simulation. Of course, “That’s not a great way to power your computer system,” admits Matt Reilly, SiCortex’s co-founder and chief engineer. “The first thing we found out was that you have to cool the people pedaling the bikes. A really good bicyclist can sustain something like 300 watts, but normally they’re moving through the air while they do that. These guys were sweating like pigs.”

Reilly and co-founders Jud Leonard (now CTO) and John Mucci (a board member and the longtime CEO) came up with the basic idea for SiCortex’s fast but energy-efficient hardware back in 2002. The time needed to finish a computation, Reilly explained to me, is usually determined by three factors: the time required to do arithmetic in the CPU, the time required to move data around in memory, and the time required for input/output operations (that is, getting data into and out of the CPU). For parallel computers—which most of today’s high-performance computers are—there’s also a fourth factor: the communications time, or the time needed to move data between processors.

Semiconductor manufacturers have done an amazing job of speeding up both CPUs and memory chips over the last three decades (but at a high energy cost, as already mentioned). I/O operations are a still a bottleneck, though a variety of tricks exist for speeding them up. But Reilly, Leonard, and Mucci—all veterans of the famed Boston minicomputer company Digital Equipment Corporation—noted that nobody was really working on the fourth problem: reducing the travel time between processors in parallel machines. “That created an opportunity for a very small company to do very large things,” says Reilly.

In a machine with thousands of processors, you can’t simply string an Ethernet cable from each processor to every neighbor that it might need to communicate with. (Imagine how many phone lines would be coming out of your house if you needed a dedicated line to connect with every home or office you might want to dial.) To keep the number of wires manageable, a parallel machine’s “backplane” or communications mesh has to take the form of a … Next Page »

Comments (1) | Reprints | Share:          

Proving that the blogosphere isn’t all about Paris Hilton or even Perez Hilton, Waltham, MA-based Interactive Supercomputing this week launched a new multi-author blog on high-performance computing called Parallel Lounge. With contributors including MIT computer scientist Alan Edelman and Cray Research veteran Steve Reinhardt, Interactive Supercomputing’s vice president of joint research, it’s the only Boston blog where you are likely to find such gems as:

Compare element-wise multiplication in a for-loop

for idx=1:n
y(idx) = 2*x(idx);
end

to a vectorized multiplication

y = 2*x;

In the for-loop, a command is sent from client to server in each iteration. This totals 1000 client-server calls, slowing down the code. Not only is the vectorized code simpler, but it requires only one client-server call.

Comments | Reprints | Share: