18 May 2011
Intel ‘blinks’ as it claims processor road map is ‘inadequate’
ARM processors were once used primarily in mobile phones. Intel's processors, meanwhile, powered pretty much every pc.
Then the world changed; smartphones appeared, tablet computers started to be developed and industry began to focus more closely on energy efficiency. Users needed different kinds of processor.
ARM had seen this coming and developed a range of cores under the Cortex umbrella. The A, M and R families – aimed at applications, microcontrollers and real time, respectively – became attractive to a wider range of applications.
ARM's low power credentials – developed from working with mobile phone companies – served it well; particularly in the burgeoning portable device market. Intel's processors, meanwhile, remained relatively power hungry.
Intel responded with Atom and one result has been the Intel/ARM face off moving into the embedded sector.
While Intel has never admitted – not openly, anyway – that ARM was getting in its face, it has announced – to an analyst meeting – that it is speeding the Atom development process. And, if reports are to be believed, Intel's ceo Paul Otellini said this was down to its processor road map being 'inadequate'.
Power consumption will be a headline feature: future devices will be designed to consume 15W, while Atom processors will be expected to consume no more than 5W.
Intel has, famously, run its business on the 'tick tock' basis as a way of matching the demands of Moore's Law. In the first year of a Moore's Law cycle, the 'tick', it delivers new process technology. In the second, the 'tock', it upgrades its microarchitecture.
A couple of years ago, observers wondered if Intel's 'tick tock' approach was slowing. Today, it looks like a new battery has been put in the clock, because it looks like it wants to move three process nodes ahead in just two years; 14nm parts are planned to be available in 2014.
But where does Intel 'tick tock' – not to mention semiconductor manufacture – go after 14nm? There is a substantial brick wall ahead; it is broady accepted that cmos won't scale beyond 10nm – and New Electronics has been discussing this in its Technology Watch section.
While ARM and Intel might, on the surface, appear to be battling each other, the real fight is about how to create reliable transistors from a handful of atoms. And if engineers can't do that, how will they continue to develop the faster, more power efficient and, above all, less expensive devices that we have all come to expect?