"Try this, assume normal yields of about 145 DPW and explain to us how AMD only produced 7.7 million processors per quarter from 2 fabs with 10,000 WSPW capacity."
Remember, there are two products, Athlons and Durons. In addition, they are pretty much segregated in their fabs. So let's look at Fab30. It likely produced all of the Athlons sold in Q3, and that was 3.8 million units. Now those chips were started in Q2. Q2 started with a maximum number of WSW of 3125 and ended the quarter with 3750 for an average of 3438, always assuming a straight linear ramp. So 3438*13 weeks gives 44,694 maximum wafer starts during that period. Oops, looks like I made a math error earlier. Still, using you figure of 145 DPW, gives 6.4 million units or so. But that is a maximum. As others have pointed out when Dan3 does his "AMD will have the capacity to make 50% of the processors in the market next year" schtick, there will be machines down for maintenance and AMD was making the transition to the AthlonXP. So 3.8 million actual units instead of a theoretical maximum of 6.4 million is the number we see. I dunno what the typical downtime of the scanners and other equipment, nor the downtime it takes to change over to a new generation of chips. Regardless, the yield cannot be as low as Paul is claiming.
As far as Fab25, that fab is heavily under-utilized. It produced only 3.9 million or so units with a smaller die and much greater capacity (roughly 5.5k WSW). Hmm, maybe this is why they want to convert it to flash... |
