X
Tech

Intel discusses upcoming laptop and Atom chips, why we need more performance in mobile devices

This week I'm at the Semicon West show in San Francisco. Most of this is inside baseball--the show is devoted to the companies that make the equipment used to manufacture chips--but in his opening keynote, Anand Chandrasekher, who heads up Intel's Ultra Mobility Group, discussed some details of the company's future mobile chips and demonstrated a few prototypes.
Written by John Morris, Contributor

This week I'm at the Semicon West show in San Francisco. Most of this is inside baseball--the show is devoted to the companies that make the equipment used to manufacture chips--but in his opening keynote, Anand Chandrasekher, who heads up Intel's Ultra Mobility Group, discussed some details of the company's future mobile chips and demonstrated a few prototypes. Much of this has been shown before at recent Intel Developer Forums and other conferences, but Chandrasekher tied it all together and made a persuasive case for continued innovation to push the performance of laptops, netbooks and smartphones.

It's no secret that it has been a tough year for the chip industry. To put it in context, Chandrasekher noted that Intel started working on Centrino nearly a decade ago in the midst of the dot-bomb downturn when desktops still dominated and for the first time PC sales declined sequentially. The iPod was also born during that downturn. In fact, he argued, many key technology innovations including the cell phone, IBM PC an even the World Wide Web had their origins in downturns.

"Every recession period, there was a fundamental invention that changed all of our lives," Chandrasekher said. "We're in the middle of this next cycle of difficult investment period. These tend to be periods where companies look hard at their investments and the bets that they make during this period usually tend to have long-lasting impact."

So what is Intel betting on this time around? More mobile, or specifically the "convergence of communications and computers." Chandrasekher repeated Intel's oft-cited prediction of a billion mobile connected devices by 2015. This growth is driven by the Internet, which has evolved over the last decade from search to shopping to social networking. Every 24 hours, there are 10,000 new Web sites and 140 Facebook applications, he said. If recent history is any indication, five years from now a list of the top sites in the world will include several new names, including sites from China (baidu.com is already on that list), India and Latin America. Mobile data traffic will double by the end of 2012, primarily because of video streaming, though other areas such as data, audio and P2P are all growing.

You'd think that most of this growth was fueled by cell phones--unit sales of handsets dwarf PCs--but Intel of course says it's all about notebooks, and that the billion devices in 2015 will be more like PCs than today's cell phones. "Yes mobile Internet growth is happening today, but almost all of it is happening on a notebook computer," Chandrasekher said. Apple probably wouldn't agree with that statement, and there's no denying that relatively expensive smartphones are growing at a rapid rate even in the teeth of a recession. But generally speaking, I'd agree with Intel's contention that limitations in the performance, screens size, network bandwidth and application compatibility are still holding cell phones back.

Chandrasekher talked about what Intel is doing to address these limitations starting with efforts to stay on Moore's Law, which is critical to increase the performance and reduce the power consumption of notebooks and other mobile devices. "That sounds easy, but it's not," he said. "The 45nm node and the high-k metal gate innovation were 10 years in the making." The Westmere 32nm processors will use Intel's second-generation HKMG, and will increase performance by 20 percent while reducing gate leakage (power consumption) by 10 times, he said. These technology "shrinks" are, of course, what enables thinner and lighter notebooks such as Acer Timeline series notebooks.

To make the case for more performance, Intel showed a Calpella laptop with a 32nm processor running a brain scan application from a company called Vital Imaging. This application takes 10 minutes to run on current Intel laptops, but the next-generation can complete the scan in only two minutes, according to Intel. "EMTs can take this on the road and do imaging on the fly in an ambulance," Chandrasekher said. "At the point of care delivery, the caregiver will be able to make decision 5X faster." Calpella, the next iteration of Intel's Centrino mobile platform, is due late this year, initially with 45nm Clarksfield processors based on the Nehalem architecture, followed by 32nm Arrandale chips sometime in the first half of 2010.

The next version of Intel's Atom processor, known as Moorestown, will still be based on 45nm manufacturing technology, but it has a new design. Previously, Intel had said Moorestown would reduce power consumption by 10X, but in May the company announced that it would actually reduce power consumption by 50 times. Though Intel isn't giving any numbers yet--most likely to avoid direct comparisons with ARM processors--Chandrasekher showed a demo comparing standby power of the current Atom with a Moorestown Customer Develoment Kit (CDK), as well as some data on power consumption when watching video or listening to music. "Trust me, it's a 50X reduction in standby power," he said. A Moorestown device will be capable of playing music for 100 hours before recharging, according to Intel.

Intel showed several prototype Moorestown devices. One prototype, from a mobile company in Finland, looked a lot like an iPhone, but Chandrasekher said it had about the same performance as notebook PC circa 2004 with the power consumption of today's smartphones. To illustrate the performance, he also showed a Compal prototype playing a 3D game, Quake Open Arena, and playing 720p video while performing other tasks. Moorestown will be "in production shortly, but it's not yet in production," Chandrasekher said. "It's just the early silicon that we're putting through its paces."

Finally, Chandrasekher pointed out that it will take more than technology to make mobile convergence happen. It will also require huge manufacturing scale "because these are not small markets." Intel has promised to spend $7 billion over the next couple of years in four manufacturing fabs in Oregon, Arizona and New Mexico to crank out 32nm processors. Those investments are already well underway, he said.

Editorial standards