What has been anticipated as the most important event in the memory industry is the upcoming court case between Rambus and Micron/Hyundai in Germany, which was supposed to start this week (Feb 16). However, due to some procedural issues raised by the defendants, the judge has postponed the trial for the patent issues for a later time. Rambus has claimed that these two memory makers (and Infineon, who will be tried in a separate U.S. case in March) have infringed upon certain patents used to manufacture SDRAM and DDR modules. With DRDRAM having very little penetration in the market, the potential revenues are significant for Rambus, and with the DRAM market being so slow, the royalties are significant for memory makers. Hyundai and Micron have counter-sued Rambus as well, apparently in the same court.
Many knowledgeable observers have indicated that Rambus will most likely win the rights to the patent, but other issues are under contention, which may affect Rambus’ rights to collect royalties. One of these issues is the claim that Rambus failed to disclose their pending patents while a member of JEDEC, a time where they were consulting for memory manufacturers as well. Since JEDEC is primarily concerned with open standards, the claim is that these technologies would never have been approved as a standard had the committee known about the patents and Rambus’ intent to charge royalties for them. One possible scenario mentioned to me by a patent lawyer not associated with the case would be that Rambus gets the IP, but must license manufacturers who were members of JEDEC when the standard was approved without royalties for several years to allow the development of a non-infringing technology. CNET ran a story recently discussing these and other issues.
In some related actions, Infineon dropped their counter-claim in U.S. court that Rambus had illegally licensed customers for technology that Infineon held patents on. Also, two of Hyundai’s charges against Rambus were upheld in U.S. federal court when the judge ruled against Rambus, but a third charge of conspiracy was dropped. Micron also has leveled charges against Rambus, which will be heard in May.
Finally, EETimes has reported on a patent case that may have some profound implications on the industry. Essentially, this ruling has limited the scope of patent claims to include only those devices specifically named in the patent application, thereby negating the ‘rule of equivalents’ which has allowed companies to charge royalties on any device that seems to remotely resemble the original patent. If the ruling stands, it could affect a number of royalty agreements currently in effect and prevent companies from altering patent applications after filing, limiting the ability of companies to extend the reach of their original patents.
The Pentium 4 created an unexpected demand for DRDRAM memory over the past several months, as Samsung found themselves running their production lines at full capacity. Combined with the planned aggressive ramp of the processor over the next 6 months, it looks like demand will continue to increase for awhile. In response to this, Samsung, Toshiba and NEC have announced plans to increase their production up to three times their current volume. It is important to keep in mind that this amounts to perhaps 20 million chips per month, which is still a fairly small percentage of the DRAM market.
What is attractive to these memory makers, however, is that the margins on DRDRAM are much better than SDRAM, which is currently selling well below manufacturing costs. Depending upon how well the P4 is accepted, other memory makers may also decide to get in on this market.
Unfortunately, if production does not keep up with demand, prices will once again rise, creating the same problem that Intel and Rambus faced last year. If P4 sales are not as great as expected, inventories of DRDRAM could rise, causing the same downward pricing pressure that SDRAM is experiencing and discouraging other manufacturers from switching over. With the economic climate as it is, pricing is a very sensitive issue for both manufacturers and consumers.
VIA technologies held a DDR summit in January, in an apparent attempt to keep it in the limelight until it finally appears in the market. The reason for the lengthy ‘beta’ period has never been fully addressed, but there have been several suggestions. One is that AMD has been waiting for VIA to have quantities of chipsets available that can take advantage of the higher FSB, another is that AMD was waiting until Micron could fix their motherboard issues, and yet another is that various compatibility problems have forced the delay.
In speaking to some individuals close to the situation, it is apparent that there have been a few timing and compatibility issues to work out, but those were resolved early this year, if not before. In addition, pricing has been mentioned as a potential problem, because those offering the limited amount of PC2100 DDR modules are charging a substantial premium over SDRAM, however several manufacturer spokespersons have assured me that this will not be the case once volume production begins. There is only a very small die-size penalty over SDRAM, and all manufacturers are using their most advanced processes with 256Mb and even 512Mb densities, making them even cheaper to make than 64Mb and 128Mb SDRAM chips, which are the current volume production parts.
At this time, all indications point to a rollout of DDR parts across the board at the end of February, or early March, since the possible reasons for the delay given above have been resolved, for all practical purposes. Several manufacturer’s modules have passed validation testing being performed by Smart Modular Technologies, and many others are in the process.
Information on MRAM from Motorola, and FRAM from Fujistu and Ramtron, was revealed at the ISSCC conference in January. FRAM (Ferroelectric RAM) is intended to be a possible replacement for Flash RAM, which is used for BIOS chips as well as digital cameras, PDAs and similar devices. MRAM stands for magnetoresistive RAM, which stores data in a manner similar to how hard disk drives do.
MRAM is being looked to as a possible alternative to not only Flash and FRAM, but main memory DRAM as well. It is relatively fast, and will hold data even with the power off. Commercial MRAM parts are not expected on the market until 2003. More information on these devices can be found on EE Times here and here
SDRAM prices continued their downward spiral in January, with little relief in sight. OEMs, who are the main buyers of DRAM memory, seem to be feeling the effects of the economic slowdown to a greater degree than system integrators and the DIYmarket. This, in turn, has created quite a glut of SDRAM chips causing prices to drop to the lowest levels in history. 64Mb chips are selling for well under $3.00 on the spot market, with 128Mb chips going for under $5.00. At these levels the small manufacturers are not likely to last much longer, and even the larger companies are struggling to make a profit. It should not be surprising to see some manufacturers focusing on more profitable parts, such as DRDRAM, Flash, specialized SRAM and other such devices.
Be the first to discuss this article!