The actual story behind intc executive adoption of rmbs over (some) engineers' questions was well known in the industry. Some just chose to ignore it.
As to the reason for this choice, who could ever be sure, but if we apply deductive logic and a bit of reasoning to it, one obvious conjecture is probably not too far from the truth :
competitive advantage over control of direction of memory technology.
However, Carl, unlike you, I don't think drdram will ever be obliterated completely. Let me try to be on the other side of the fence this time. <g>
1. There are still niche applications for which drdram appear to be quite suitable. Frame buffer in DFP displays with analog connector. This is not an insignificant niche. The price of TFT panels have crashed recently, with 15" going for <<$500 and 17"<900, the replacement of CRTs with LCD seems to enter the realm of possibility now.
2. There is some indication that drdram sales has gone up significantly recently, as the P4 sales return closer to projected numbers. Percentage wise at least (if not by absolute numbers)
3. The graphics guys may have solved DDR usage well, but the P4 mobo chipset guys have not proven themselves yet. If they screw up the first launch, DDR adoption could be delayed significantly.
4. To match the P4 quad-pumped 100MHz FSB, the DDR NorthBridge using 2x64 ddr @2x100MHz(old terminology, memory controller hub(?)) needs to build some sort of data packer/unpacker, this will add a clock or two to the latency. The granularity of data may also go up to 256 bytes, which will cause DDR some of the same problems that plauged drdram. (ie: performance increase over sdr may be negligible).
SbH |
