After 28 years of experience at Intel, Pat Gelsinger is a veteran of the age of the microprocessor, with a CV that dates back to work on the 286 chip and continues today with his role as senior vice president in the firm’s Digital Enterprise Group.
That role has Gelsinger overseeing the bulk of Intel’s processor and platform development. With Penryn, Intel is now pushing yet another new line, and trying to counter the perception that hardware performance is now outpacing what users need to run software.
Gelsinger denied that processing power has outgrown desktop application needs. “The question [at the time of the 386 launch] was the same at the time as we get now,” he said. “Moore’s Law is reasonably linear. We’re doubling performance of our chips approximately every two years, but software follows hardware. The task is to create a vacuum and performance jumps up to make that the new normal.”
As for what will fill that vacuum, Gelsinger said that the answer is always unpredictable, as no one can tell what new application might be just around the corner. “There was a time when Mitch Kapor wasn’t Mitch Kapor”, Gelsinger said, in a reference to the Lotus founder who created one of the first killer apps for the PC, the Lotus 1-2-3 spreadsheet.
When pressed, Gelsinger indicated two areas where more raw speed is still needed. The first is graphics rendering, where he believes that “polygonal representations are soon going to disappear” and fields such as global illumination and ray tracing, both for photo realistic lighting effects, will become mainstream soon.
Faster processing will help show “how cloth acts when it’s moving or how
fabric stretches across a frame… the behaviour of objects interacting with each
other so that when a ball hits a window, you can model the behaviour of breaking
glass”.
The broader area of visualisation is also hungry for CPU power and Intel has a
particular focus on healthcare applications such as real-time rendering of
magnetic resonance image (MRI) scans.
Speech recognition is frequently cited as another application that chews through processor cycles, but Gelsinger sees a broader opportunity to create an interface capable of interpreting not just speech input but also watching users to see what gestures they make.
“If you look at some of the hard problems out there, they require levels of performance that we can’t yet build,” he said.
However, the challenge for chip makers has been made more complicated in recent years by the requirement to drive down energy consumption, and Gelsinger still hopes to see the day when laptops are capable of “all-day computing”. This would mean running for at least 10 hours, even with Wi-Fi on, a situation he conceded is still several years away.
