1) No, I don't think it will be the IBM 90nm process that fab36 uses, it will hopefully be the AMD/IBM 65nm process running at 90nm. Hopefully there are some synergies from the combined experiences of both AMD and IBM. Best app/form/design/process should rule no matter who came up with the idea. Fab30 is all 200mm although the requirements of 300mm production were well known at the time. So it is possible that fab30 could be converted to 300 mm should it be needed. When fab30 was developed 300mm was bleeding edge, now it is mainstream.
2) The last thing I read was samples 4q05 and production H106. Realistically, I doubt if fab36 will be ready for equipment before the end of the year. It's one thing to get the walls up it's another thing to get the inside finished. Look at how a house is constructed. It might take a week or two to frame it, but a couple of months to finish the insides. By that schedule they would start bringing in the machinery about the 1st of next year.
3) Personally, I want to see some IBM beef. Considering the number of delays IBM has had with Fishkill I really don't know how much faith we should have in their Processes. Which is one of the reasons why I'm very interested in how AMD's 90nm process is going.
4) Most of the 65nm equipment should be backwards compatible. I would expect AMD to bring up the first line on 90nm just to limit the number of variables. If that proves acceptable (binsplits/volume) it could provide any extra product needed and reduce the pressure to get to 65nm. In any case it would provide a fallback position in case 65nm has unexpected problems.
5) It remains unclear just what AMD plans are with respect to covering a production gap that should be starting to hurt by q205. Right now the only possible alternate production site that I can see is Fishkill. Maybe they could move the k7 production there, but I'm hoping that AMD's 90nm process works well enough and that the synergies developed were enough that Fishkill can handle K7 and K8 production. This again would reduce the pressure to get fab36 up and running and provide some breathing room.
6) In any case AMD can't just throw a switch and start producing double fab30s volume. The markets need to be greased for that sort of production increase. I doubt that fab36 will see full production until 2007/08 or later. A game plan is needed that allows for increases in production as demand warrants.
7) If fab36 comes up faster than expected and the costs of production are such that it makes sense to move all production to fab36 then fab30 could either be updated to 300mm for processor production or be used to make flash, or some combination. In any case AMD is going to be able to produce an awful lot of something.
News 
Market Data 
Discover 
