This week CNET reported that Intel (NASDAQ:INTC) will develop another Atom architecture, one built on Intel’s upcoming 22nm process. The implications of this piece of news are vast and it shows that Intel is at least talking seriously about taking on ARM.
Atom isn’t the very best x86 chip in the market for PCs. AMD’s Bobcat outperforms the current iterations in both performance and power consumption, at least on a system level. This has more or less put Intel out of the Netbook market for the remainder of 2011, and forced them to drop prices tremendously. Atom is not an architecture intended to take the bottom of the PC market, it is an architecture intended to take Intel to the mobile space, which means tablets and smartphones. Atom is not to go up the ladder, it was meant to go down.
Until now the road to the mobile market has not been without hurdles. When they initially developed Atom they thought that 45nm chips would be enough to enter in the smartphone market, but in the meantime, two things happened: First, ARM processors made huge leaps in performance and at the same time they kept a very tight power envelope. Second, the mobile market changed, nobody wants their mobile phone to run Windows XP or Windows 7. The mobile market demands very different solutions when compared to the PC market.
By failing to beat ARM in power consumption, Intel simply did not have the means to get design wins and start shipping phones with its chips. Even worse, ARM chips are reaching a point where they can power small netbooks such as the Asus EEE Transformer, and servers in specific applications. This means that not only is the netbook is threatened, but also the very lucrative Xeon Business.
Today Intel said it will address that, they will change Atom’s strategy. They will implement a Tic-Toc cadence for Atom, which means a new architecture every two years. They will try to beat ARM in the power curve by keeping Atom one node ahead in the manufacturing process at the same time retaining a performance advantage and x86 compatibility. There is a lot of room to play in a 22nm chip, but Intel has an even larger set of challenges. Given their track record so far with Atom, they now have to prove their capabilities, not just talk. At the moment, Intel is only talking.
If Intel manages to deliver what they promised, they will finally have an actually competitive part, not just slides. This may slow the adoption of ARM chips in the netbook space and in the server market by eating part of the market and diminishing the returns of ARM manufacturers. The real question is whether 2013 is just late or too late.S|A