Hi Dave B; RDRAM and RIMMs are (mass produced) components, not finished products. The high multiples are for finished (non commodity) products. The context of our conversation was pricing of products containing embedded processors and memory, not pricing of components containing memory.
Embedded computers with memory subsystems generally don't go into commodity goods, they are used in specialized products. Specialized products sell at high COGS ratios.
An example of an embedded computer that was sold for years into a relatively low markup commodity product was the Zilog embedded processor that used to be contained in nearly all mice. The chip was about $.25 each in large quantity, and was the most expensive part of the whole product.
I know that you can analyze the component costs of a product and see that small changes in the cost of a single component results in a small change in the profit margin on the finished good, but that is not how industry (in my experience) decides on retail pricing. Instead, the engineers calculate a COGS, and manufacturing applies a 3.5 or 4 or even higher multiple on that aggregate cost in order to compute the retail price. That number, along with the specifications for the product are given to sales and marketing, and they estimate how many they could sell at that price per year. (They are always wildly optimistic.) Then company management decides whether to put the product into production or not.
It's actually a bit more complicated than this. Companies try to figure out how much it will cost to build the product well into the future. The costs usually drop, as memory prices, for instance, decrease. This is why the memory makers refusal to guarantee lower prices for RDRAM in the future reduced design wins. When SDRAM started out, some of the memory makers quickly guaranteed (contract) pricing parity for SDRAM vs DRAM early on. While the spot prices were undoubtedly all over the map, the contract pricing was what companies used to figure out cost of goods sold. Instead, with RDRAM, some of the memory makers have repeatedly stated that the stuff is expensive and difficult to manufacture. That is not the kind of thing that engineers like to hear when designing in new products. You can call it FUD from companies that are getting sued by Rambus, but it predates the SDRAM / DDR suits, and even if it is FUD, it still has an effect on design engineers. If you design in a component that your supplier tells you is going to carry a heavy premium, and then it doesn't come down in price, you look stupid later on.
-- Carl
P.S. As far as Sony claiming that Rambus saved them $40 per unit, that is BS. What do you expect Sony to say, that Rambus costs them money, but they are including it anyway cause they're stupid? What happened is that Sony designed in RDRAM many, many years ago, long before Intel gave hints of supporting alternative technologies. Now they're stuck, so they are putting the best face on it that they can. |
