Forgot to post the conclusion. Here it is. However, I suggest all read the whole thing for a better appreciation of DDR's death. ;-)
intel.com
The PC industry has recognized that a new memory technology is needed to allow processor performance to continue to improve. Although there has been significant discussion on the various merits of DDR and RDRAM technology, there’s no question that neither technology fits completely into the current SDRAM infrastructure.
Implementation of any new memory system requires extensive development, simulation and verification along with new manufacturing processes and procedures for all affected elements of the PC supply chain, from semiconductors to systems.
This work has already been completed for RDRAM, and will continue to pay dividends as the inherent headroom in the technology allows the performance to scale both in frequency and width over the coming years.
DDR is much less far along the development path, with commercial samples only becoming available recently. This is nearly 18 months after the first RDRAM systems were publicly demonstrated by Intel at Comdex ’98.
Just as importantly, no organization has stepped forward to lead and fund the technology validation required in order to ensure the system robustness and vendor interoperability which is essential when shipping millions of platforms using components from dozens of vendors as Intel has done with RDRAM.
Finally, even should all of these essential tasks be completed, the underlying technology cannot be extended further without substantial changes approximating what is needed to get DDR to market, all over again.
The need is here for a desktop memory technology to support GHz class processors and the new internet applications. RDRAM systems are validated, stable, deployed, and ramping in volume, with sufficient headroom to carry systems through the next several generations of processors.
DDR is Burst EDO all over again. A seductive idea, but without the sufficient technical advantages over SDRAM needed to make it compelling, even while requiring an entirely new infrastructure. |
