During the recent Platform Conference it was said by participants that massive DDR-II SDRAM-based platforms appearance will take place in 2005, a year later than it was originally scheduled in the past. There are loads of reasons to hold the new standard from becoming mainstream sooner than in two years time and it seems that the main factor to delay the DDR-II is slow personal computer market in general and mass availability of PC3200 (DDR400) to fill the gap between DDR and DDR-II technologies. A Micron representative Mr. Brett Williams said that system manufacturers has not indicated that they are ready to move to DDR-II as soon as possible and they seem to be the main driving force of such transition.

Major DRAM manufacturers have been sampling DDR-II SDRAM memory chips since Spring 2002, but we should consider the fact that to build computer systems with a new type of memory requires to validate designs of mainboards, chipsets, memory modules and a lot of other devices, what is not cheap at all. It is said that Samsung currently achieves nearly 40% yield of DDR SDRAM at 400MHz, so PC3200 (DDR400) is still far from being cheap, thus, we may expect that the yields of more complex DDR-II devices should be lower and the cost of final chips higher. Given that the market is slow these days and vendors have dual-channel PC3200 (DDR400) memory for their needs, there is no sense to force transition to DDR-II, especially if there are difficulties with qualification of the technology.

Even developers of servers now validate PC3200 (DDR400) memory for their needs. According to Silicon Strategies ServerWorks had told conference participants that it would sample its Grand Champion SLX chipset in the first half of this year to support PC3200 (DDR400). David Dorrough, ServerWorks' technical marketing manager, had added that a ServerWorks chipset supporting DDR-II SDRAM was to enter production in the second half of 2005.

PS. In case we are going to stick with 6.4GB/s PSBs till 2005, or, maybe very late 2004, I wonder for what reason did SiS develop its 4-channel RDRAM R659 with up to 9.6GB/s of peak memory bandwidth? Even if Intel had plans to introduce 1066 or 1200MHz Quad Pumped Bus for future microprocessors, they now either have to reconsider the intentions or implement exotic technologies like QBM SDRAM or RDRAM in future chipsets. Frankly speaking the latter is unlikely.