To: BUGGI-WO who wrote (202210 ) 6/15/2006 8:26:47 PM From: Joe NYC Read Replies (1) | Respond to of 275872 Buggi, Correct, it will be difficult to supply 14-15Mio. units in Q3 and 10% more in Q4 with 2-3Mio. DC (Q3) and hopefully ;-) 4-5Mio. in Q4. I think we can assume that Q1 had 13.5M units with about .95M units X2 (7%). We can also assume that AMD was running at capacity. To go from here to Q2, let's say that shipment rate in Q2 is down about 10% (flat to slightly down = -3%, 14 week quarter = -7%). Let's suppose that ASP stays the same (declining SC prices offset by richer X2 mix), the unit reduction should be 1.35M units. If the entire reduction is in SC processors (very likely, there would be some extra capacity, 1.35M worth of SC chips. Now, roughly speaking, if this 1.35M worth of SC chips is applied at another 1.35 SC chips making them DC, this is what I have: Total SC DC Q1 13.50 12.55 0.95 Q2 12.15 9.85 2.30 This should possible entirely without any additional capacity. Now I will try to plug in your Q3 and Q4 estimates (high end DC, low end unit), we can see how much additional capacity would be needed: Total SC DC DC% Q1 Joe 13.50 12.55 0.95 7% Q2 Joe 12.15 9.85 2.30 18% -------------------------------------------- Q2 vs Q1 -10.00% -21.51% +142.10% . . Q2 Joe 12.15 9.85 2.30 18% Q3 Buggi 14.00 11.00 3.00 21% -------------------------------------------- Q3 vs Q2 +15.22% +11.67% +30.43% . . Q2 Joe 12.15 9.85 2.30 18% Q4 Buggi 15.50 10.50 5.00 32% -------------------------------------------- Q3 vs Q2 +27.57% +6.59% +117.39% Maybe I am a little high on Q2 being 18% of total. I think your 32% DC in Q4 is right on the money. Can someone refresh my memory on how much capacity can be expected from Chartered and Fab 36 for Q3 and Q4? Looking at the numbers of what will be needed, I don't see AMD being capacity constrained. I was not counting on any contribution of Fab 36 in Q2, while there will obviously be some. I think that AMD should be replenishing the inventory in Q2 to move back to normal, and in Q3 and Q4, the ramp of Fab 36 should roughly follow the seasonal increases + SC to DC conversion. Or better, meaning more than seasonal increases + expected SC to DC conversion. What I think the reason for elimination of 2x1MB CPUs is not capacity, but cost. Last Q I figured that AMD blended cost is about $38 per CPU. That may translate to, say $30 SC and $60 DC, which would be for maily 512K L2 chips. If 2x1MB DC part costs $80 to produce, the economics start to get ugly, as the prices these can be sold start to approach $100. As long as premium for DC remained high and percentage of DC of total CPUs produced was low, the high cost (based on my estimates) did not matter at all from cost standpoint, only from capacity standpoint. As capacity grows, prices drop, the high cost and lack of 65nm is bound to hurt AMD badly on gross margins. 300mm is on the way to reduce the pain, but only 65nm can remove the pain alltogether. Joe