Fusion next on DAAMITs agenda

[W1nTry]

Bulldozer is off the roadmaps for now

Replaced by Swift

By Charlie Demerjian: Thursday, 20 December 2007, 9:57 AM

AT THE RECENT AMD analyst day, people noticed that Bulldozer was off the roadmaps and replaced by Fusion. To the twits out there, this was the end of the world, AMD is done, stick a fork in them. They are wrong.

Don't misunderstand, this isn't good news for Bulldozer (BD) and Bobcat at all, nor is it any sign that Fusion, specifically Swift is a stopgap or hack. Both are two completely different products aimed at completely different markets.

As long ago as last summer, word was floating that AMD was getting into a scheduling crunch. The problem is that they went from a handful of development teams to many more, something around two to seven, but it depends on how you count. Some of the problems stem from that, others from other areas.

The biggest crunch is that both Fusion/Swift and Bulldozer were scheduled to come out around the same time, BD was set to tape out next September, and Swift around then as well. AMD didn't have the resources to do both at once. More than six months ago, there were fierce internal arguments over which one would go first.

The thinking was that BD would win back server marketshare, something that they needed badly, and Swift would do the same for enterprise notebooks. Both are aimed at big, lucrative markets where AMD has a low marketshare. Both had very credible arguments as to why they should go first.

We know who won, that is of course Swift, the key question is why. The answer to that is obvious, BD is a mess right now, and it is delayed. How much? Damn good question, that is the part that will make or break AMD. If it is a quarter or two, not the end of the world. If it is 2011 like the most dire predictions I am hearing, well, don't put away that fork just yet.

BD is a quite radical departure from what you are used to seeing in an x86 chip, and if it works, big if currently, AMD has the potential for a great product. The hard left turn in chip design is course a potential minefield, and that is what led to Swift winning as much as the delays.

In the end, there wasn't enough to do both at the same time. Swift is a fairly safe bet, basically Griffin Mark 2++, done with proven (by then) components. Low risk. BD is high risk, late and potentially a non-starter. The internal fight turned out to be a non-issue.

Given that, AMD did right. They still have time to put BD back on track and get things going. Maybe they will, maybe they won't, but for the next year, all we have is the results of the decision to make Swift first. µ