Arm wrestling with Microsoft tests Intel's muscle

Rupert Goodwins: Intel and the handheld makers want us to buy into the XScale processor. But Microsoft's not playing ball, and perhaps it has a point.

Answer the following question in ten words or fewer, and win a job for life as an Intel marketing manager: "I need a faster processor in my handheld computer because..."

Not easy, is it? But if you found it insultingly simple -- and anything mentioning video games, streamed media or fast encrypted communications will get you on the plane to Santa Clara -- then try the advanced question: "And Microsoft should create an optimised version of Pocket PC for faster processors because..."

This is, as far as can be told, an unanswerable question, and it's one that's causing acid flashbacks at Intel. Remember the Pentium 4, with its astonishing new architecture that went much faster than the older chip if you ran optimised software? Remember how long it took for any optimised software to appear, during which time the chip looked like a dog, barked like a dog and ran like a toddler's nose? Intel remembers. In fact, there's a strong argument that its new XScale chip should really be code-named the DejaVu. It's got an astonishing new architecture that goes much faster than the older StrongArm if you run optimised software -- but there's no optimised software. As the benchmarks show, it's turning in results that are sporadically excellent, mostly so-so and occasionally horrific when it runs the old code.

What Intel wants Microsoft to do is produce an XScale-optimised version of the Pocket PC operating system, so all that zoomy newness is unleashed. What Microsoft wants to do is pretend XScale doesn't exist: hey, with a name like that it might easily be a toilet cleaner or an anti-dandruff shampoo. Microsoft's attitude is understandable enough. It knows very well that developing and maintaining multiple different versions of the same operating system is expensive and error-prone. It used to have many different versions of Windows CE for many different chip architectures; it even had several versions of NT. It knows that abandoning multi-platform compatibility has done it no harm at all, it knows that it has the whip hand in the relationship with Intel and it doesn't mind letting the chip company know it.

Furthermore, it knows that the number of licences it will fail to sell if it doesn't create an XScale version of its Pocket PC operating system is precisely zero. All those new XScale PDAs will still run the non-optimised version of the software just fine: it's not as if there's anything else for them to do. They'll be slower than they might be, but the day Microsoft starts worrying about how fast its software runs is the day that there's some competition. In the Pocket PC arena, there's none. Palm software lives in a parallel universe where people just don't run big applications, and even the most rabid Linux weenie will giggle if you suggest it's ready for prime-time retail on handhelds. Where there is competition is on the chip side, where Intel faces TI, Motorola and others with ARM-based processors that will be just as nippy in real life.

Which leaves Intel with a number of alternatives. It can -- and does -- encourage application software writers to optimise for XScale, as there are definite competitive advantages here. It can jolly along network and video driver writers, write some codecs itself, covering as many bases as it can to get software speeding up without duplicating the operating system. It can even do what it did with the Pentium 4, and just keep upping the clock frequency until the inefficiencies are wiped out by pure muscle: expensive, especially for the PDA makers who have to play keepy-uppy, but Intel's got a fair amount of clout itself.

But this leaves a lot that's not going to get beefed up any time soon: filing system, memory and application management, user interfaces and input/output are going to remain the remit of Microsoft. That's OK, says Intel, these aren't areas where a faster processor makes much difference. But as these areas cover the vast majority of what you end up doing with a PDA, that's tantamount to admitting that you may not need a faster processor after all.

That's as close to heresy as you can get in this godless business of ours. Nevertheless, look for the real excitement in handhelds: bigger storage, faster communication and more beautiful displays. What the processor is won't make a huge amount of difference: something Intel will be trying hard to have us forget. Expect a big Intel Inside splurge for PDAs, don't expect the benchmarks to back it up, at least at first. Hey, California's not such a great place to work just at the moment anyhow.