To: Skeeter Bug who wrote (47783 ) 8/24/1999 11:42:00 AM From: Carl R. Respond to of 53903
Both. Graham is talking about cost changes with feature size, and I was talking about cost changes by choice of product given a feature size. Manufacturers do keep moving to new smaller feature sizes because there are ultimately cost savings, as Graham points out. But given that you are going to produce at say .21µ, what product should they produce? 128mb? 64mb? 256mb? or 16mb as Stefan suggests? Or even 1gb chips? (Yes, MU could make 1gb chips today if they wanted to, but the die sizes would be huge, and the costs prohibitive.) Obviously what you want is the highest profit margin. Typically there is a premium for the highest end product, but the low end products all sell for roughly the same cost per megabyte. Thus the answer to the most profitable product at the low end depends on which is the cheapest to produce per mb. At the high end the question is which goes up faster, cost or price? In making a chip there are several different costs. First is the cost to produce the die, and there the costs rise at a higher rate than linear, which always favors smaller chips, which was my prior point. There are also costs which are approximately the same regardless of size, such as handling, packaging, shipping, selling etc. If these costs were the only factors you would always produce the largest possible chip. As for testing costs and repairing costs, I'm not sure but I suspect they are more closely related to the size of the chip and therefore neither favor large chips nor small chips. How these different factors interplay affects your decision of which chip to produce. The fixed handling, packaging, and selling costs seem to make it impractical to sell chips under $4-5 each, so in answer to Stefan's question, yes the die cost per mb for 16mb chips would be lower than 64mb, but the overall cost per mb would make that product an inefficient choice. At the other end, if you are looking at 256mb and 1gb chips the fixed handling and packaging costs are pretty much irrelevant, and the question is whether the die costs rise faster than the selling premium. These costs can play out differently at different companies, by the way. If one company is more efficient than others in the area of handling, packaging, and selling, then they are likely to stick with a prior generation longer because for them the savings in die cost is sufficient to keep the low end product a better choice. Similarly a company with a higher fixed cost per chip will move to the next generation sooner. I for one doubt that at NEC the cost to produce a 128mb chip is currently less than the cost to produce two 64mb chips. I think that NEC made the move expecting a price premium, which is no longer there. Yes, at some point the cost savings on packaging a 128mb chip will offset the die cost penalty, just as at some point the cost and packaging savings at 256mb will offset the die cost penalties at that size. Carl