Rupert Goodwins' Diary

Tuesday 6/6/2006Intel continues to have baffling ideas. It is, as mentioned on Monday,an x86 world, with the amount of non-x86 processing – on the desktop,at least – rapidly approaching zero.

Tuesday 6/6/2006

Intel continues to have baffling ideas. It is, as mentioned on Monday, an x86 world, with the amount of non-x86 processing – on the desktop, at least – rapidly approaching zero. So the news that Intel is planning to ditch some of its non-x86 architectures should come as no surprise. After all, despite a lot of effort, there's still no sign that the Itanium will make any money any time soon, right?

Only it's not the Itanium they're rumoured to be ready to cut adrift, according to reports in the Wall Street Journal. It's the embedded stuff – the Xscale and network processors. Now, these are perfectly good designs which have taken a lot of the useful components out of the x86 and linked them in to an ARM core. They work, they're nice and low power, but they haven't been fulfilling expectations. Why this is, I don't know – the embedded market is quite ferocious and has a number of established players, and Intel came late to the game. That may be enough in itself: it's very difficult to dislodge people with existing relationships to OEMs. And the embedded world has been more prone than most to dodgy dealings of the "If you want to buy this chip, you have to buy this other one" variety – which can lock people into families of components for a long time.

None of this changes the game Intel's beancounters are playing, which is undoubtedly ROI based. This business costs us this much and makes this much in return – is that the best use of our money? Can we assume useful growth? This sort of thinking, looking at yourself as if you were an external investor, is why companies divest themselves of profitable businesses: it's not a matter of making money, it's of beating the market. Hence IBM getting out of PCs, and hence Intel getting rid of parts of the company that look to the outsider like they're doing perfectly well where they are.

Financially astute this might be, but it takes little or no account of the people involved. Intel has a unique culture that inspires a lot of loyalty, and that translates to a willingness to go the extra mile when it matters. Of course, there are always disappointments in any job – projects get cancelled, bosses resign, divisions get reorganised – but throwing away a whole market segment with everyone attached is a bit like lopping off a limb. The shockwaves can affect all manner of people for a long time, especially among those who feel most strongly attached to the company.

But does anyone care?