Intel restructures chip architecture groups

ABOUT SIX WEEKS ago, and around about the time that Craig Barrett's memo lit up the 80,000+ world of Intel Inside, a decision was taken at Intel to re-organise the way the firm designed its IA microprocessors.

While the details remain shadowy, and it's unclear what jobs will be affected, the move underlines Intel's determination to rip up the old roadmaps and start all over again. It's an attempt to impose consistency on an organisation with too many irons in the fire, it appears to the INQUIRER.

While CEO Craig Barrett's memo stressed the need for better "execution", this is a little bit rich, since the decision of the captain of the Good Ship Intel to turn round the oil tanker 180 degrees is a little late in coming.

We reported from the Spring Developer Forum that essentially the Pentium 4 is a dead duck, and we believe that the vague talk about dual cores at the Fall Developer Forum last week is a holding operation while the klaxons are klaxoning inside the mighty operation. In fact, the lack of detail about the nature of future dual core technology underlines the intense fight still going on inside Intel.

In contrast, AMD's story at the moment on its architecture seems to be us to be consistent and logical. It's not forecasting a major sea change to infrastructure for a move to dual core technology. Intel's story about dual cores at the desktop, notebook and server level is full of inconsistencies and there's more question marks than there are answers.

We should not read too much into the absence of Louis Burns from his scheduled appearance at the show last week - after making inquiries it appears that personal matters prevented him from banging the usual desktop gong.

But it is also quite clear that if, as we were told, a broad re-alignment is taking place across chip design units, there's a bitter war between the former divisions as to which unit gets the upper hand, Louis Burns' or Anand Chandrasekher's. Our money, right now, is on the latter's unit.

It could be that Intel has already formed a steering committee of high fliers to make the changes across the chip architecture divisions. What's pretty clear from last week's Developer Forum is that we're seeing a new, cautious Intel which doesn't want to pre-announce technology which it may have problems delivering.

It's pretty clear that Intel - from 2001 onwards - has tried to do too many things and in that process taken its eye off the ball. Now it appears to be taking steps to change things, but at the same time Intel is going through a handover period at the top. Craig Barrett is due to retire in the next six to nine months, while Paul Otellini is being groomed as his successor. In some ways, this makes the radical IA architecture changes now underway harder, rather than easier. µ