Introduction
According to Sherry Garber and Bob Merritt from Semico Research Corp. the memory market will see quite some diversification over the next two years. This is mainly due to the convergence in home entertainment – broadcast programs, home computing and Internet content are merging onto one platform, and the convergence of voice, data, video and Internet in the network. Or as every keynote speaker at every networking or electronics conference keeps saying: ‘We are moving from the industrial age to the information age.’
This is good news for the memory industry because all the new emerging markets like handheld assistants, Internet appliances, advanced cellular phones and the digital home environment with set top boxes, HDTV (High Definition TV), cable and DSL modems need some type of DRAM. In 1999 the worldwide revenue of DRAM was $20 Billion and it will more than triple to $78 Billion in 2004. Semico predicts that the DRAM consumption by application is going to change in favor of the communications and consumer market and declining in the desktop and notebook segments. The forecast of revenue market share in 2004 prognosticates 57.4% DDR DRAM, 42.4% SDRAM, 0.1% RDRAM and 0.1% EDO DRAM.
The hottest gossip topic at the Platform 2000 conference was the lawsuit Rambus filed against Hitachi at the beginning of January. Apparently Rambus is charging Hitachi with taking patents that were actually intended for the design of Direct Rambus DRAM chips and using the technology for SDRAM interfaces instead. Besides punitive damages Rambus demands an injunction against almost all of Hitachi’s DRAM products and SH microprocessors with SDRAM. But this is not all: Rambus also claims that the entire DRAM industry should be slapped with the same restrictions. Needless to say this did not exactly go over well with other DRAM vendors. Rumor has it that several major DRAM companies have quietly offered their help to Hitachi by sharing data that supposedly proves the synchronous technology was developed long before Rambus filed its patent claims in 1990.
DDR-II to the rescue
The industry’s answer to Direct Rambus DRAM (RDRAM) is DDR-II. At the conference the Future DRAM Task Group of the Joint Electronic Device Engineering Circle (JEDEC) gave a talk on the evolution of DRAM. The current task of the group is to develop DDR-II DRAM and aid the JEDEC committees during the standardization process. The JEDEC committee is currently in the ballot process for the DDR-II standard, targeting the entire spec to be finished by the end of this year. If they can meet the deadline, DRAM vendors should be able to produce DDR-II in volume by 2003.
DDR-II is expected to offer a minimum bandwidth of 400 MBits/s per pin based on 100-MHz signaling. The interface is supposed to run at 1.8 Volts, down from the 2.5 Volts of DDR-I, and it will need new packaging at the chip and module levels, as well as a new data-capture and synchronization scheme. However, despite the higher pin count of DDR-II there will be a common DDR-I and DDR-II DIMM and no custom DRAM interface is required on the memory controller. One important focus is DDR-I backwards compatibility. This is for example achieved by doubling the total bandwidth of DDR-I by increasing the data fetch from 2 to 4 bits. As a result, the same 100-MHz SDRAM core will increase from 200 to 400 Mbits/s per pin with DDR-II.
But who will win the DRAM battle, DDR-II or Rambus Direct DRAM? At the conference the opinions varied since it obviously also depends on the market segments. The mainstream PC market is very cost-sensitive and Rambus’ chips are expensive. Rambus demands a licensing fee, and the technology also requires a new manufacturing infrastructure. On top of that the Rambus dies are bigger and have lower yields. For example Intel already dropped the support for RDRAM for mobile processors. According to Farhad Tabrizi, Vice President for Strategic Marketing and Product Planning at Hyundai MicroElectronics, the projected demand for RDRAM chips this year is 100 million units. But he thinks the real number will be closer to 20 to 30 million units, which is nothing compared to the entire DRAM market. Because of the high cost RDRAM it only makes sense in high-end PCs and this segment is less than 5 percent of the total market and continues to shrink.
However, some industry experts say that the superior speed of RDRAM will keep Rambus in the market. Currently RDRAM offers 800 MBits/s per pin and is supposed to double over the next few years. Rambus was suspiciously absent at the conference. One attendee commented: ‘Why should they come? All they have to offer is an inferior product at a higher price.’