InvestorsHub Logo
icon url

pgerassi

06/20/04 4:32 PM

#38410 RE: sgolds #38376

Dear Sgolds:

Then you do not know how a fab really works. You can have 90nm product and 130nm product running at the same time, like right now. Testing and validating 65nm product and processes can be done at the same time as 90nm production. It is not like a typical production line where one conveyor goes down and many machines work at only one point on that line. What moves is pods of wafers. They go to a photoresist coater, then a step exposer which exposes the photoresist, then to a process that develops the photoresist which removes the unexposed photoresist. Only the step exposer has to be 65nm and 90nm masks can be used in it. All of the other machines simply work on the wafers. Now a 90nm step exposer can be used instead of the 65nm one. They both can be present. One has 90nm masks and the other 65nm masks.

There is a finite amount of time to change masks on a step exposer (called that because it exposes the mask image onto one die area and moves the wafer to expose another area ("steps" to the next die target on the wafer)). This is the setup time. Then there is a finite amount of time it takes to expose one wafer. The amount of wafers that can be done in a hour is called the wafer run rate. This will include pod load and unload time.

So the work at a particular step exposer is scheduled by the pod. A number of events like "mask unload 14a, mask load 15b, expose pods 254, 256 and 257" are scheduled. So a single machine may expose 5 or 6 different steps on 15 to 18 different pods in any given day. The mask and the associated parameters are kept together logically. Thus, the line is really a flow path through the Fab's machines. And the 65nm validation is made by getting a good receipe for each step in the flow path for that particular process.

IMHO, 6 months of 90nm output may pay for all of AMD's portion of the Fab financing. So the last half of 2005, it can run 90nm production while the Fab is being validated for 65nm processing. And in early 2006, 65nm production can be started. Any "delay" will be an artifact of the machine scheduling. Which places it at management's feet (at least the plant scheduler).

Pete