bobs10, would it bring up 300mm faster if they started it with 90nm? It doesn't make much sense to me. Had they designed it that way from the start then it could lead to faster production. However, Fab 36 is being implemented with the 65nm process that AMD and IBM are jointly developing. Changing horses midstream usually causes a project to slip, not to get done faster!
Another thing to consider: The footprint of 300mm wafer handling equipment is larger than 200mm equipment. Fab 36 has similar size to Fab 30. Thus it will have less production lines. Although you do get twice the number of raw dies from 300mm, because the footprint of machinery is larger it does not translate into twice the production from the same floor space.
Now you may be envisioning Fab 30 and Fab 36 running at the same time so the increased production would mean that AMD could supply 30 to 40% of current CPU demand. A reasonable hypothesis. However the costs for AMD will go through the roof during the first year that 300mm is in production because you should expect low yields. Thus Fab 36 may actually have less output than Fab 30 in the first year if they are both at 90nm. Those Fab 36 processors will cost a lot to make!
Since the equipment is designed for 65nm then the only thing that makes sense is to bring it up on 65nm and take whatever lumps to get the process tuned. The same wafer will produce more usable dies on 65nm, even at a significantly lower initial yield than 90nm would have produced (on 300mm).
Besides, Hector was quoted in the article that Keith just posted as saying that Fab 36 will start production in late '05, ahead of what we were expecting (samples in late '05, production in early '06). So far, so good (although usually the schedule slips remain hidden until later in a development process).
My take: Damned the torpedos, full speed ahead on 65nm!
