Sshhh... One-chip, real-time smartphones are here
A quantum leap vendors don't want to talk about
2006 Preview By this time next year the bar for mobile technology will have been raised rather higher than it is today, in what promises to be the most significant hardware advance for several years. But in a surprising contrast to most technology introductions, which either trumpet arcane features (eg, "32 bitiness"), or are noisily hype features before they're ready for use, this is a move the manufacturers don't even want to talk about.
Even odder, is the fact that devices featuring these aren't secret. Many of you will have read about them already, and TV and web advertising campaigns have already begun in the UK and Europe for the first batch.
So what's going on here?
The breakthrough is the confluence of two trends, each long predicted, but only now deemed ready for prime-time. The first is the consolidation of functionality onto one chip. Most of today's phones (and wireless PDAs too) use two processors to perform much of the donkey work - a baseband processor to handle the radio, such as GSM, CDMA and Wi-Fi, and an application processor to run the operating system, such as Windows Mobile, PalmOS, Symbian OS or Linux.
But running everything on one chip, the application processor, allows the manufacturer to make a considerable saving on Bill of Material (BoM) costs. Or alternatively - and this is the route being taken by manufacturers with the first clutch of devices to be announced - a smartphone or PDA can be delivered packing a lot more functionality, or more expensive parts, at the same price point as the model it supersedes. We're used to Moore's Law and falling component prices taking care of business, and giving us annual improvements, but this more of a quantum leap in price/performance. TI claims a 20 per cent saving in BoM costs, which only tells part of the story.
The CPU in question is Texas Instruments' TI 2420. Thanks to its investment in DSP chips over the years TI dominates the high end phone business, but its OMAP platform has struggled in the performance stakes when matched against Intel's Xscale chips, particularly with floating point intensive operations.
The 2420 has a five-fold performance increase over the OMAP1710 used in today's high end smartphones, and around a 100 fold increase in floating point operations.
Texas Instrument describes the 2420 as a "smartphone on a chip" - but the chip is only part of the story. To run take advantage of a one chip devices requires an operating system fast enough to handle the radio interfaces, such as GSM and CDMA. Is there a real-time operating system in the house?
Here the story takes an even more curious turn. The first clutch of devices to use the 2420 are billed as Symbian OS devices, but there's no reference to this being, effectively, an entirely new operating system, the biggest overhaul of what was once known as EPOC in a decade.
Symbian's real-time kernel began as a skunkworks project in 1998, but was only released to licensees, with very little fanfare, two years ago. It's known as EKA2, for "Epoc Kernel Architecture 2", and last month John Wiley published an exhaustive introduction, clocking in at almost 1,000 pages, should you have $85 to spare. (A good, shorter introduction from the book in PDF form can be downloaded from the Wiley site here.
There's more to EKA2 than the new nanokernel approach - Symbian has added new IPC mechanisms - but the system provides low latencies that permit real-time operations required for radio operation. The EKA2 kernel is available for both version 8.0 and 9.0 cuts of Symbian OS, with devices bound for Western manufacturers opting for the latter (due to the added security features), and Japanese manufacturers, who run "closed" phones, continuing with v8.1b retro-fitted with EKA2.
Sony Ericsson's P990 smartphone, and Nokia's N series and E series are the first devices to feature EKA2 - the P990 using Philips' Nexperia application processor - and both manufacturers have chosen to increase the feature set rather than slash the BoM cost. In almost all of these devices, the manufacturer has opted for a higher density screen and Wi-Fi.
Which begs the question, why the lack of fanfare? It's all the more puzzling since Nokia's neither Microsoft nor PalmSource have a real-time mobile offering at the moment.
Take your pick from a number of reasons. BoM costs are sensitive information, and manufacturers don't like rivals to make more than guesstimates. And there are times when it's prudent to conceal major breakthrought from the competition - they might notice.
However there are also pragmatic reasons, and neither vendor wants to affect current device sales or upset its ISVs (independent software vendors). With Symbian v9, there is some pain with the gains. The new N series and E series phones, and UIQ 3, are all incompatible with existing software, although this is largely due to the Platform Security ("PlatSec") features rather than EKA2. Software will need to be recompiled.
And why now? Real-time, single-chip phones using Symbian have long been on the vendors' drawing boards, one source told us, but it wasn't economical to do until air interface standards had settled down. With 3G, EDGE and 802.11 now fairly stable, it is now prudent to introduce the features. ®