Post a Comment
Intel’s Atom came with a fight
Intel CTO Justin Rattner said there were at least three attempts–maybe more– to sell the company on the notion of building a low-power IA processor.
By Ann Steffora Mutschler, Senior Editor -- Electronic News, 6/11/2008
MOUNTAIN VIEW, CALIF -- Reflecting specific areas where Intel Corp is investing some of its annual $6 billion in research, the chip giant today demonstrated and discussed more than 70 projects and concepts in development in its labs at its annual Research@Intel event, which covers topics including visual computing, wireless, health, environment and life sciences.
Intel Senior Fellow and CTO Justin Rattner (pictured left) kicked off the event held at the Computer History Museum by calling attention to products and technologies that form the basis of the company’s direction, the latest being the Atom processor.
Rattner went on to say that he could think of at least three attempts–maybe more–to sell the company on the notion of building a low-power IA processor.
“The significant work began in 1999 within the Microprocessor Research Lab [MRL]," he explained during an opening address. "The idea was to take the P6 microarchitecture–which is basically the Pentium Pro, Pentium 2, and Pentium 3 microarchitecture–and see if we could strip it down, if we could simplify it, in order to get the power down for something that we would consider a low-power device, just a watt or two. And, in fact, there were some research results and this idea of a low power processor was reviewed with the senior staff but didn’t quite make the cut."
However, things got much more serious in 2002, Rattner said, within MRL or, as it is known today, the Microprocessor Technology Lab (MTL), one of the three big labs in Intel’s corporate technology research organization. A group working primarily out of Intel’s Austin lab created a new microarchitecture called “Snocone,” which was the basis for a low-power IA design and dramatically simpler than any of the mainstream cores including the mobile cores.
“[Snocone] was taken to sufficient degree of definition and simulation that we convinced ourselves that a core on the order of one watt could be built with a very small design and it would have reasonable performance in that it could go into applications that were typical of personal computing devices," the CTO said. "At that time, we were focused on the notion of ultra mobile PC devices so it had to be good enough, fast enough to host the range of PC applications."
With that data in hand, Rattner said a much more comprehensive effort was put together to convince management that such a processor could be built, should be built, and would open up a large array of application opportunities, not just for ultra mobile PCs, but for a range of devices where the Intel architecture was attractive but the power levels at that time were in the seven to eight watt range were not practical.
“In 2004, work started to actually build the product we now know as the Atom processor. Atom is a clear example of where long-term persistent efforts on the part of the research organization to explore an important idea and opportunity for the company have ultimately had a big payoff. The reception to Atom is tremendously gratifying to the research organization because we believe in this idea wholeheartedly and to see it received so well by the industry is satisfying,” he said.
“We’re still at work figuring out how to build smaller and faster ones so there should be a very healthy pipeline of technology going forward for the Atom family,” Rattner concluded.
