May 2001 Industry Update

Pages: 1 2 3 4

Memory

More old news is the fact that Rambus, for all practical purposes, was shot down in flames in Federal court this month. Not only were the patent infringement claims thrown out, but the patents in question were limited in their scope. Then, to add injury to insult, the jury found Rambus guilty of fraud in their dealings with JEDEC. Furthermore, just this week an Italian judge threw out all Rambus claims against Micron. The only question that seems to be left is whether Judge Payne will rule that the Rambus patents are unenforceable under the rule of ‘equitable estoppel’, which basically means that because Rambus was not dealing in good faith with JEDEC, their applicable patents cannot be enforced on any JEDEC members. This would essentially mean that SDRAM and DDR royalties are no longer an issue with regards to Rambus.

Ironically, these decisions may have had the exact opposite effect that people expected. Instead of causing manufacturers to completely abandon DRDRAM, they now appear to be more willing to embrace it. This may be due to the fact that Rambus is now less ‘scary’ as a player in the industry, and is evidenced by Intel’s recent decision to once again push back their i845B chipset (P4 DDR) several months at least. Though this may seem like a bad thing for those who are staunchly opposed to anything Rambus, it could be exactly what the industry needed to move forward. The P4 was designed for memory bandwidth, and currently only DRDRAM can deliver.

Much of the DDR vs. Rambus ‘war’ was politically motivated, rather than technically. Both technologies had their benefits and drawbacks last year, but DDR was the one that had the political advantage, for several reasons. As long as there was a threat of Rambus controlling the industry, the politics were more important than the business or technical issues. It has now become obvious that DDR is not a panacea for all memory woes, and DRDRAM does have a huge benefit in bandwidth, which the Pentium 4 was designed to exploit. With the most serious drawback for DRDRAM no longer that much of an issue (price), it can be seriously considered a contender for the next memory standard.

Going out on a bit of a limb, and being sure to state this only as a personal opinion, I believe it is possible that by the end of next year we could see DRDRAM with a fairly significant market share (more than 20%), with all memory makers producing parts, all chipset makers including support in at least one of their chipsets, and AMD making products that utilize it. My reasoning here is that something must replace SDRAM, and DDR seems to be losing momentum right now, Intel is still strongly behind DRDRAM, SDRAM/DDR margins are non-existent, and since Rambus has been brought down to earth the resistance from memory makers has a chance of crumbling. No doubt Micron, Infineon and Hyundai will not be extremely happy if this occurs, but on the other hand, Rambus would not be in much of a position to deny a license to manufacture, since it could be their only source of revenue.

Before the pro-Rambus crowd starts claiming that Rambus has ‘won’ (and anti-Rambus people start flaming me), it is only a suggested possibility based upon conversations with a few well-connected people, and the fact is that in this fast moving industry technologies are replaced very quickly. Whether DRDRAM or DDR is the next memory standard, it would likely only be a two or three year reign, since there are already several potential replacement technologies making their appearance, including FRAM (Ferro-electric RAM) and MRAM (Magneto-resistive RAM).

Currently, approximately 75% of all DRAMs are used in computers, according to Dataquest, however this is expected to decrease to about 50% in a few years. Embedded applications are becoming more popular, and graphics applications are growing as well. An advantage for MRAM is that it not only is fast (3ns access vs. 60ns for DRAM), but is non-volatile and requires less than 1% of the power DRAM does, . In fact, Electronic News posted an article a few weeks back that states “MRAM appears to offer a route to the ultimate all-purpose semiconductor memory—fast to read, fast to write, nonvolatile, durable and dense—something the chip industry has been seeking for 30 years.” Of course, we’ve all heard similar hype about certain memory technologies before…


Pages: « Prev  1 2 3 4  

Be the first to discuss this article!