ARM vs. Intel? It's ARM for now

LONDON — Processor intellectual property licensor ARM Holdings has been working with partners for 18 months to define a product category to sit in the gap between the smart phone and the laptop computer. Meanwhile, the world's largest chip maker, Intel Corp., reckons it knows what will sit there and has been talking about the Ultra Mobile Personal Computer for so long that it calls the platform the UMPC. And in 2007, Apple launched the iPhone.

Who has the better strategy for mobile processor success? Will future mobile handhelds run on an X86 or ARM instruction set architecture?

Intel's strategy is to provide X86 processors that will "offer leading performance while reducing the footprint and power consumption," according to Jon Jadersten, European marketing manager for the ultra mobile group at Intel (Santa Clara, Calif.). In other words, Intel wants to equip the X86 to invade ARM's traditional domain: low-power handhelds. To that end, it has produced three "platforms"--McCaslin, Menlow and Moorestown--essentially reference designs that show OEMs what is possible with upcoming Intel silicon.

ARM's strategy, meanwhile, is to keep producing processor architectures and cores optimized for the semiconductor manufacturing capabilities of its licensees and the needs of their customers.

Intel's Menlow consists of a Silverthorne 45-nanometer processor, a support chip called Poulsbo for controlling I/O and graphics, and a communications module that can be either Wi-Fi- or WiMax-capable. Moorestown combines the functionality of at least the first two chips into one, and Intel claims to be reducing idle power consumption by an order of magnitude with each platform.

But neither platform intrinsically supports GPS, 3G or even 2G.

Jadersten said young people will want to continue the social networking activities started on the PC and take them mobile, and he predicted this form of interactivity will dominate handheld traffic for the "always on" generation. He argued in favor of the downward flow of software, saying, "The software community is already on X86. They can develop on the PC and deploy on Menlow."

But not everyone is convinced. Indeed, Francois Meunier, semiconductor analyst with JP Morgan Cazenove in London, is inclined to suggest a pox on both their houses.

Score one more design win for ARM: Warren East, CEO of the IP licensor, tries out a TyTN II mobile Internet device made by HTC.

"There's a lot happening at both the high and the low end," said Meunier. "Nokia wants the low-end phone to have access to the Internet because it wants to expand service revenues. ARM may be the incumbent, but are we going to see five ARM cores in future designs? No."

Still, he said, "the mobile phone is better positioned than the mobile PC to fill the vacuum. If you want to know why, ask consumers. Mobile phones don't crash. Intel needs to question its form factor."

Intel's already doing so. It has begun talking about Mobile Internet Devices (MIDs), which would sit below the UMPC but above the smart phone, and look a lot like the iPhone. "The Internet experience on the handheld is suboptimal. We believe we can close that gap," said Jadersten. While early 2008 introductions will be UMPCs as Menlow gives way to Moorestown, the MID will come to dominate, he said.

But for some observers, what Intel is talking about is already here--and powered by ARM.

"We're talking about screen sizes in the 5- to 6-inch space," said Ian Drew, segment marketing vice president for ARM (Cambridge, England). "We see it [the product category] as a grownup smart phone. We certainly think our experience of more than 10 years of saving power in mobile phones is relevant; it's what allows a variety of products."