A blanket that monitors your baby's heart rate, a farm crop that requests more water, a swimming pool that calls for help. All of these fantastic-sounding products will share a common bond: they'll all have Intel inside.
Chip giant lays out plans for future
At least that's the plan of Intel vice president and chief technology officer Pat Gelsinger, who described these and other far-reaching ideas during a keynote yesterday at the Intel Developer's Forum in San Francisco.
In addition to meeting Moore's famous law, Intel intends to expand upon it, Gelsinger said. Instead of just pushing for more and more performance, Intel will also drive its technology into new areas — bringing processing power and intelligence to products that have never used it before.
Among Intel's grandest plans — in the next five years, the company wants to put an entire radio in the corner of one of its CPUs. It won't be just part of a radio, but the whole thing — antenna and all, Gelsinger said.
"We want to make radios so cheap they're ubiquitous," he said. Ubiquitous radios that are dynamically configurable will make it possible to create a seamless network across offices, buildings and even cities, he said.
Intel could make this work through an ad-hoc sensor net, Gelsinger said. To demonstrate the concept, he brought out David Cullen, director of Intel's Research lab at the University of California at Berkeley.
Cullen led a demonstration of his lab's Ad-hoc Sensor Net by having the audience reach under their seats, where more than 100 of them found boxes about the size of TV remote controls. He activated the network, waking the sleeping nodes which then linked with each other to create a stable network.
Then Cullen introduced three more nodes, each attached to beach balls. It took a few moments, but these bouncing nodes eventually joined the network, too.
Technologies like this make it possible to have a sensor-based baby blanket, Gelsinger said. Another use: placed throughout a building, these nodes could awake and notify engineers if structural problems arise.
Today these sensors are the size of a 10p piece. Soon, they'll be a mere speck of silicon, Gelsinger said.
Looking toward future products is fun, but Intel knows it must address basic production issues just to keep pace with Moore's Law, Gelsinger said.
In the beginning, the challenge was to innovate fast enough. It took Intel three years to go from a 25MHz processor to a 50MHz processor, "and we were proud", Gelsinger said. Today the company sees jumps like that every day.
Currently the company's biggest problem is power and heat, he said. On its present path, Intel's high-speed chips will eventually generate the heat of roughly "the surface of the sun", he joked.
In order to keep pushing processing performance, Intel has to find ways to lower the power requirements, he said. One possible solution is Intel's hyper-threading technology, which simulates the work of multiple processors in one, increasing performance by as much as 30 percent, he said. Another is the use of higher-performance circuits that use less power.
The simple fact is, Intel can't continue to increase the power to its chips. "We're not going to 10,000 watts," he said.
Gelsinger also addressed Intel's plans to bring silicon together with optical technologies. The combination could greatly improve performance of fibre optic networks and other products, while driving down overall costs, he said.
And all these plans revolve around the expansion of Moore's Law, Gelsinger said.
That 'law' is something Intel Chairman Emeritus Gordon Moore — in a videotaped segment — said he "blindly extrapolated", but which "turned out to be more precise" than expected.
Looking forward, Gelsinger said, Intel is more confident than ever that it can maintain at least another decade of attaining Moore's Law.
"Moore's Law: we see no end," he said.