The number of computing engines on microprocessor will increase
rapidly, with hundreds appearing on a single chip, Intel's top
technology expert has predicted.
Patrick Gelsinger, general manager of Intel's digital enterprise
group, said Intel was preparing to develop chips built around
hundreds of computing engines, rather than two to
four in current chip designs.
Chips like these could help in medical applications and
tomography which requires 3D models of patients. Speeding the
rendering time could speed diagnoses, he said.
Office-based applications like speech recognition could replace
typing input and use the improved processing power to deliver more
interactive software to users. "It becomes immersive, intuitive,"
said Gelsinger.
However, for businesses to take advantage of many-core computing
they will need to develop new programming languages to interface
with more complex processors.
Gelsinger said the miniaturisation of chips for
devices such as mobile phones would also bring PC-style software to
more mobile devices.
Intels’ x86 chip architecture will increasingly be
pushed into cellphones and consumer products. The biggest obstacle
in designing chips for mobiles is their
high power consumption, he said.
Intel will have to work fast in the mobile market.
Apple said last month it would use technology from PA Semi to make
chips for the iPhone. Gelsinger was disappointed with the decision
but hoped to win back business from Apple. “People haven’t been
lining up outside. We have to go earn this business,” he said.
Intel will develop chips with a focus on internet
connectivity. Its aim is to give every person on the planet
internet access 24 hours a day, Gelsinger said. “You will be
interacting with the computing experience without ever thinking of
it,” he said.
Intel is moving to use a new generation of the
silicon wafers which would cut the average cost of producing each
chip by 40%, said Gelsinger.
Chip companies unable to spend at least $1 billion
on designing new types of chips would be unable to keep pace with
Moore’s law, which states that the capabilities of chips double
every year or two.
“Our task now before us is to crack terascale
computing at everybody’s desktop level,” he said. “It’s not going
to happen with a broad range of programmers. We need a new language
to abstract those problems from the users.”