Hi wily; There aren't any cost advantages for RDRAM in the chart you linked. The number of pins in the memory controller has very little to do with total system cost. There is only one memory controller, and 40 pins isn't going to make a big difference. In contrast, there are typically at least 16 memory chips, (much more for big machines), and the RDRAM cost adder (royalties, silicon, and c.) applies to each of them.
This is typical Intel BS. They're comparing RDRAM to PC1600, rather than PC2100, so you figure they're comparing for the big systems rather than the desktop, right? But they only show 2 sockets for DDR, while there are plenty of shipping products with 4 DDR DIMM sockets. In other words, they're making the desktop assumption when they count the number of DIMM sockets, while at the same time making the server assumption when they count bandwidth.
Re: "Also, one could look at Intel's plans for DDR P4 support as just a way to keep the RDRAM manufacturers honest -- as it is, they have Intel over a barrel with P4 requiring RDRAM. They also need the DDR chipset for server platforms and for the value segment."
One could look at it that way, LOL, but then how do you explain AMD, VIA, ALi, SiS, ServerWorks, Micron and Nvidia's failures to design RDRAM chipsets? No, the truth is that everyone in the industry knows damn well that RDRAM is dead, dead, dead (for new designs), and they're busily converting over to DDR production. The marketing people at companies that need to push RDRAM chips (because they were late to convert over to DDR) are the only ones trumpeting the technology, to the rest of the industry it is long dead. Even Samsung is pushing DDR.
-- Carl |
