200mm vs 300mm Costs:
1. Obviously you have more consumables per wafer. This includes items like Raw Wafer Cost, Photoresist, gases etc. Thus consumables will be 2.25x, but this impact should be small, since consumables are not significant.
2. Any 300m tool will cost more than 200mm tool. WAG - 30% extra. Thus depreciation will be 30% extra.
3. Labor and overhead cost will be the same per wafer.
So your WAG estimate for wafer cost may not be that bad, and the cost per square inch for 300m wafer will be cheaper than 200m wafer.
However I will be surprised if the percentage yields are the same for the first year. Larger wafer will have fewer percent losses due to partial die at the edges. However the process control will inevitably be worse across the larger wafer, particularly since all the 300m equipment is not mature. For example even on 200m wafers yields on the edge are poorer, because of poorer process control on the periphery due to equipment issues. However since the periphery consists of partial die anyway, the equipment vendors stop refining the equipment to achieve better process control. Suddenly as the wafer size is increased, the equipment limitation becomes apparent. They finally fix the problem but it can take a year or two of refining the equipment design. So I expect the percent yield on the 200m core of the 300mm wafer to be equivalent to the 200mm wafer, but a significantly lower yield for the remainder( outer core) of the wafer. All pieces of equipment do not suffer from this problem. However a silicon wafer has hundreds of process steps, and it takes just one process step which is poorly understood by the vendor or the process engineer, for the yield to be less than expected.
My pessimism stems from having seen 5 wafer size transitions from the days of 2 inch wafers. I hope that Intel does not suffer from the same pitfalls. |
