Replies to post #38254 on Advanced Micro Devices Inc (AMD)

[CombJelly]

"y'all seem to agree that it would be easy to bring up 90nm on a production line designed for 65nm"

No, just pointing out that AMD could be leaving that option available. If nothing else, they are looking for at least one engineer to do 90nm design work at Fishkill at 90nm for, presumably, fabbing out certain designs. Given that Fab36 will more than likely use the same equipment(assuming IBM uses the same equipment for 65nm as 90nm) due to the joint development of 65nm, then this scenario is possible.

Given that we don't know what equipment will be used at 65nm, this is just speculation. But it isn't in the "highly unlikely" category.

[bobs10]

By the time fab36 comes online money will be a lot less important than it is in today. It had better be or there are going to be a lot quite disappointed AMD investors.

Hopefully by then AMD will be so supply constrained that getting product out the door will be more important than bringing up 65nm.

Transitioning capacity demand to 65nm by using IBMs fabs is probably a good idea if it can be done so both parties make money. Even with just one line fab36 is going to add a ton of chips that will need to be sold no matter whether it's on 65nm or 90nm.

It should be interesting to see how INTC responds to the fab36 weggie.

[pgerassi]

Dear Sgolds:

IBM's Fishkill fab is a 90nm SOI 300mm fab. And it is being used by both IBM and AMD to get a 65nm SOI 300mm process ready. Is there much doubt that AMD is also using that fab to iron out the 90nm SOI 300mm processes to be used in fab 36? Thus fab 36 can produce 90nm SOI 300mm output much quicker as the starting tests are already complete. Thus, it is probable that fab 36 could get 90nm production by Q3 2005. This is just in time for the 90nm K8 dual core dies to be produced. Using this low volume output to qualify fab 36 is another fine idea.

BTW, isn't this another variation on Intel's own "copy exact" fab expansion policy? There may even be 90nm dual core Opterons produced at Fishkill, first as samples than later as production needs arise. That would make IBM into a foundry for 90nm and later, 65nm AMD64 CPUs. And AMD has a knack of making foundries profitable. Just look at the flash fabs.

Of course another way is that AMD buys the Fishkill fab from IBM or becomes a joint owner like it is with Spansion.

Pete

[jhalada]

sgolds,

Even though you get the support for 90nm built in to the 65nm equipment there is still a design cycle to adapt the parts produced in Fab 30 to the new process in Fab 36, plus the testing and the ramp on the new line. In product development you can have huge testing and bringup costs even when there are very minor engineering changes.

Yes, but part of the preparation can be done at Fishkill. If AMD does indeed sign foundry agreement with IBM, there may be actual production parts in Fishkill, moving it to Fab36 should then be simpler.

Joe