Future Fabbing . . .

In the AMD conference call earlier this week (BTW, both Intel and AMD held up pretty well in round one of the price war), there was some newish revelations/confirmations about AMD’s fab plans.

  • The new fab, Fab 36, is at 50% of full production capacity now, and Fab 36 will be the place where the first (and most) 65nm chips will be made.
  • Chartered will start converting from 90nm to 65nm capacity in 2007.
  • By the time 65nm production constitutes most of Fab 36’s production (and Chartered starts adding to that), Fab 30 will begin to phase out (starting in the second quarter of 2007). It looks like they’ll take some production off-line convert to 300nm use, and once they get that going (conversion will happen during 2008), then they’ll pull the rest to convert that. Fab 30 turning-into-38 will continue to have no less than 40% of maximum production during the conversion.
  • Fab 30 will be rechristened Fab 38, and it will end up with the same production capacity as Fab 36.
  • 45nm production will begin starting in mid-2008 at both fabs.
  • What about the New York fab? Well, it’s still in “maybe we will, maybe we won’t” mode.

    Thoughts? Well, for the first time, one can see how AMD plans to get to 100 million processors by 2009 (being able to sell them at a good price is another story for another day), but process shrinks will be key to increasing production.

    AMD is going to have to ramp up 65nm production in 2007 even if they can get only mediocre speeds from it, first to simply accomodate a much higher percentage of dual-cores being made, then to allow Fab 30 to start going offline.

    If that gets delayed (or if Chartered has some problems with its conversion), that’s going to create real headaches the second half of 2007, because they’ll have to either put out fewer processors, or delay the conversion of Fab 30 to continue to make what will be pretty oldish 90nm chips.

    It’s unclear whether Fab 38 will make some 65nm chips for a short time in 2008, then go to 45nm. The big question for 2008 will be “Will quad-cores catch on?” 2007 isn’t a problem, Intel estimates that a year from now, only 3% of its desktop CPU production will be quad.

    If quads become mainstream in 2008, though, AMD will be hard-pressed to hit 100 million CPU capacity even with 45nm capacity, and if there’s delays, they’ll be in the same disadvantageous position they are now making 90nm duallies vs. Intel’s 65nm chips. . . . .

    Ed


  • Be the first to comment

    Leave a Reply