BOISE • Imagine a computer that would search for mutations in complicated strands of DNA in hours rather than days.
Or a cell phone that could in moments scan all its stored videos and cull out those with images of Grandma by using facial-recognition technology.
Such possibilities and more may come from a new processor from Micron Technology Inc.
The Automata processor, announced by Micron at the recent Supercomputing 13 conference in Denver, is being called a revolutionary approach to processing massive amounts of data while enhancing speed and efficiency. Its applications range from medicine to security to sifting and analyzing complicated and difficult
Automata processors also may enhance the company’s fortunes. It provides Micron yet another path toward product diversity and away from its reliance on cranking out basic memory chips, such as dynamic random access memory (DRAM).
Announcement of the device comes six months after Micron bought bankrupt Japanese chip-maker Elpida Memory Inc., in a move that industry analysts called a steal for the Boise company. The purchase contributed to Micron’s $1.71 billion profit in its fourth quarter, up from $43 million the quarter before.
People are also reading…
It also comes just months after the company announced a 5 percent reduction in its 30,000-employee worldwide workforce following Elpida’s purchase, including, according to the Idaho Department of Labor, more than 350 at the Boise campus.
Company officials won’t give specifics on how research on products like the Automata could affect jobs at Micron. The company employs a little more than 5,000 people at its Boise campus.
While Micron’s new processor is generating positive buzz in the industry, it’s coming of age with some challenges.
“It’s a relatively novel technology that needs to break into some well established markets that do things other ways,” said Brian Shirley, Micron vice president of DRAM Solutions. “It still has a sizable path in front of it to get to the point where it would be millions of devices.”
A New Highway
Automata (pronounced ah-TAH-muh-tah) is a reconfigurable device with tens of thousands of tiny processing elements fusing logic and memory, said Paul Dlugosch, director of Automata processor technology. The device can crunch tens of thousands of bits of information at the same time.
It’s like a freeway with thousands of lanes to handle traffic, compared to conventional processors that have fewer lanes and a fixed regimen for processing, said Jim Handy, founder of Objective Analysis, a California company that watches the semiconductor industry.
The device can process and analyze unstructured data, which doesn’t fit into a neat, confined database. And its thousands of tiny processing elements give it an edge in pattern recognition.
Micron’s new processor holds potential for detecting corporate cyber attacks, even when the manner of the attack — images, files, word-processing documents — is unknown. Or it could look for similarities between two different patients who exhibit common symptoms, said Dlugosch.
“This is a big change,” Handy said.
Years in the Making
Yet in some ways, the ideas behind the Automata processor isn’t new. It’s based on a decades-old theory about data processing that many thought would never become reality.
A team of 20 people at Micron spent seven years developing the chip at the company’s research-and-development hub in Boise.
Micron kept financing the research through some of its most difficult years, even as it laid off 3,500 people and closed a fab at the 2008-2009 height of the recession. Work continued while the company sank into nearly three years of red ink through 2009, as the volatile memory market pummeled Micron.
Developing the Automata processor was a “swing-for-the-fences” proposition, Shirley said: “We are a firm believer that to survive long term you have to be an innovation leader.”
Company officials won’t say how much they spent developing the processor. “We were very pleased with the overall low development costs and are very excited about the marketing opportunities,” Shirley said.
Micron’s new processor hasn’t left the company’s property yet. Researchers are still finessing the device. By mid 2014, Micron hopes to make the Automata processor available to researchers, some universities and companies to try on a limited basis.
Asking New Questions
Micron isn’t discussing a date for when full-scale production and sales will begin. But the company has helped finance the new Center for Automata Computing at the University of Virginia for research into possible applications for the processor. The university put up $300,000 for the next two years; company and school officials declined to say how much Micron contributed.
While Micron does have a plant in Virginia, the project with the University of Virginia largely grew out of a previous relationship Micron had with university faculty, said Kevin Skadron, center director and chair of the university’s Computer Science Department.
The center is designed help Micron’s processor break into the marketplace by bringing together researchers and key industries to examine the processor’s potential, said Skadron.
Skadron sees rich opportunities in the processor’s ability to efficiently analyze voluminous amounts of data. It could lead to new avenues of research — where crunching data was “so expensive, you never asked the question,” he said.
His center will look at medical applications, such as analyzing so-called “big data” like images of human cells. But the Automata processor also could be put to work sifting complex data on economics or national security.
Skadron echoes Shirley’s admonition that the Automata process won’t be a quick sell.
“People are reluctant to invest in learning a new programming model and taking existing programming and rewriting (it),” he said.
Speciality vs. Commodity Products
For Micron, the processor is another avenue to develop specific products for specific purposes, as opposed to manufacturing and distributing generic memory chips — the key to the company’s business for years. Memory chips are a commodity; Micron makes a profit by keeping production costs below sales costs. But that hasn’t always happened in a market known for wild swings in supply and demand.
The Automata processor aims at specific customers, much like the company’s long-term relationship with car companies to design specific memory chips for automobile systems — a move that gave the company increased margins.
Handy, the semiconductor analyst, said a prime customer for Micron’s new processor could be the National Security Administration, which gathers huge amounts of data — such as millions of phone calls — in need of analyzing.
Whoever the customers are, they will be getting a specialized product to fit their needs, said Shirley, the Micron VP. That may mean they won’t sell as many as they would memory chips, he said. But they would be of “much higher value and therefore more profitable.”