Alan,
I'm glad I couldn't finish earlier. The article from Investmill posted by REH adds a lot with respect to why DDR DRAM won't succeed. I've copied the section on DDR DRAM below.
But going back the instability question, I think I'll stick primarily with the previous answer (which, granted, wasn't much of an answer other than to say that both DRDRAM and DDR DRAM are facing similar problems). I think that this is backed up by Investmill's points #1 and #2 below. DDRDRAM is not a slam-dunk technically--it is pushing the limits of an old technology. Point #2 also supports the fact that there was a change in the SDRAM specs when the move was made to 100Mhz by saying that it is incompatible with 66Mhz SDRAM. I'm assuming that Herman's explanation of Intel having to jump in to assure that everything worked was correct.
As I've often said, I'm investing because Intel is supporting the technology (see #3 below) and because consortiums rarely produce break-through technologies (see #4 below). I expect that both DDR DRAM and Direct Rambus DRAM will eventually work but that Rambus will win the medium to long-term battle. You gotta believe that if Intel, Sony, Nintendo, et. al. have selected it, they've done so for a reason. I'm sure they've had hours and hours of discussions with Rambus technoids.
FWIW, the "elegant" part of the Rambus technology that Herman likes is the fact that the clock makes a round trip in the Rambus memory -- he says it helps remove some of the timing issues. I decided to go check out the Rambus literature to understand this a little better and quickly got out of my depth. I'm going to sit down sometime with Herman and get him to explain it to me in detail.
Dave
p.s. Alan, I know that this doesn't even begin to answer your specific question about DDR DRAM. I'll see what else I can find out. But keep in mind I'm a marketing guy who happens to have a technology background and who enjoys technology and can understand most of it (and the rest when it's explained), but it isn't my primary role. Somehow I have to lose this reputation for being the technology guy on the thread <g>.
Or get better at it. <ggg>
My natural inclination is to do that latter.
p.p.s. I know there are some techheads reading this board. C'mon out folks and jump in to the discussion!
___________________________
Competition from DDR
InvestMill web site
Several DRAM manufacturers who are opposed to the idea of investing in new equipment and then paying royalties to Rambus are proposing a competing standard called Double Data Rate (DDR). DDR is an evolution from current SDRAM that was developed by the Joint Electronic Device Engineering Council, and it is a royalty-free open standard. DDR basically doubles the data rate between the CPU and memory to 200 MHz. The open standard has allowed DDR supporters to push the technology further, and now several DDR proponents are claiming bandwidth speeds approaching that of Rambus, such as IBM claiming a 1.6 GB/s version and Hitachi claiming a 2.1 GB/s version.
Thus, the industry is divided into two competing camps: DDR supporters consisting of memory manufacturers and RDRAM supporters consisting of microprocessor makers. Of the two, we believe the advantages of RDRAM clearly outweigh the DDR. DDR offers little more than a low-cost transition and stronger near-term supply due to easy conversion for the DRAM makers. Meanwhile, Rambus offers many important advantages, as outlined below:
1.The potential bandwidth growth for RDRAM appears much greater than that of DDR. DDR bandwidth is limited by bus width that has met CPU limitations, and limited by data rate increases by pushing the limits of an old technology. In contrast, RDRAM can add significantly more channels to increase bus width and can increase data rate in future generations.
2.The improvements in DDR may not be as simple as the DDR supporters would like everyone to believe. Remember, this is an industry that has only produced three significant upgrades in memory technology during the past twenty years. Furthermore, it is important to remember that 100 MHz SDRAM is incompatible with 66 MHz SDRAM, opening the question for system board manufacturers that they may be forced to completely redesign the system with every DDR upgrade.
3.RDRAM is supported by the CPU manufacturers compared to DDR supported by the DRAM makers. CPU manufacturers wield considerably more influence in the market, and they have been unresponsive to DDR. Of the major manufacturers, only Compaq's Alpha has offered firm support for the DDR standard, even though it has voiced the same support for RDRAM. AMD has been silent and Intel has been vocally opposed. In contrast, all the CPU makers have supported RDRAM as the standard in future processors and all the top 15 DRAM manufacturers have announced plans to produce RDRAM.
4.We have doubts about the long-term potential of the DDR coalition. One DDR coalition (the SLDRAM) has already collapsed, and the others are already competing among themselves to offer the best version of DDR. Plus, the idea of cooperation among such commodity-like competitors reminds us of OPEC, where cooperation has proven all but futile. Initial RDRAM production equates to greater profitability for the initial manufacturers, while DDR offers little more profitability than available today for SDRAM. We believe the support for DDR among the manufacturers will quickly evaporate as competitors begin shipping RDRAM and realize greater profitability.
In short, the Rambus standard offers greater potential, fewer questions, and stronger support. Regardless, the DDR standard will likely make a dent in initial Rambus acceptance, and will probably cause significant fluctuations in the stock price as DDR supporters make claims that may or may not be proven true. We believe DDR is simply a distraction to the obvious long-term acceptance of the Rambus standard. |
