When Intel announced that they would be backing the Rambus DRAM standard, there was widespread industry concern that this was merely attempt by Intel to control the memory market. At the time, it was assumed that any proprietary implementation was doomed to fail, and that another ‘open’ design would win out in the marketplace. Since that time, the Rambus technology has gained momentum and now appears to be the favorite for future implementations. Despite concerns by many analysts about potential drawbacks and issues, there have been no complaints or comments from manufacturers about implementation issues or problems. There seems now to be little concern by the general public and we see almost no mention of competing technologies, as most have conceded the war to Intel.
One of the reasons that we have heard very few negative reports about DRDRAM is due to the licensing practices that have been implemented by Rambus and Intel. In order to become privvy to the technical details necessary to even truly analyze the technology, a manufacturer must ante up a large sum of money. This ‘buy in’ merely allows the manufacturer into the game, with additional royalties to be paid along the way. The DRAM manufacturer must pay a royalty, as must the module manufacturer. In order to use DRDRAM, the motherboard manufacturer must also license the ASICs and RIMM slots, which reportedly carry a hefty price tag. However, this isn’t really the major problem. Apparently, in addition to charging these royalties, Rambus and Intel have forced these various players to sign a contract with prohibits them from publicly making *any* negative comments about the technology. If they violate this ‘rule’, their rights to the technology can be revoked!
Recently, there has been some reports of dissention within the ranks of the Rambus Elite. The manufacturers who bought into the technology are reportedly concerned about the issues and problems which have yet to be addressed by Rambus and Intel. Few details are available because of the ‘gag order’ in effect, but these manufacturers are saying that without the ability to publicly criticize the technology, Rambus and Intel have very little incentive to address the issues. It would seem that rather than a very efficient, well designed memory technology what we have is an extremely controlled, and somewhat questionable marketing tactic, a la Microsoft!
Both Rambus and Intel have been touting this technology as being the best option available. If this is so, why would they need to restrict any negative comments about it. Their licenses don’t appear to restrict positive comments, so it would seem that they believe that this technology cannot withstand a head-to-head comparision with competing technologies that are cheaper and outside of their control. This should be a very big concern for users and manufacturers alike.
Intel already has incredible control over the direction of PC processors and chipsets, and now are trying to do the same with memory. While many believe that this is a benefit to consumers, there is evidence to show that this is not the case. For example, with Microsoft dominating the operating system market, they have had no incentive to make the Windows 9x operating system more reliable or compatible with more hardware. If there were another OS manufacturer that could run Windows applications, Microsoft would be forced to work on stability rather than just add features to drive others out of the market. What is truly disappointing is that so many consumers accept this situation because they have never experienced a truly stable computer system. Frequent crashes and other problems are considered ‘inevitable’ and must be tolerated, though those of us who have been working with business level machines for decades know this to be untrue.
The danger with the Rambus/Intel licensing practice is that this may now be the case for hardware as well. The big reason that Rambus technology is touted as ‘better’ than other solutions is because of the greater bandwidth, however for the majority of users bandwidth is not an issue. Though the Socket 7 platform benefitted from the 100MHz bus speed, it was not because of the higher bandwidth from memory, but because of the improved bandwidth from L2 cache. The Pentium II processors proved this by moving L2 cache off of the memory bus, so that this platform gains almost nothing from the increased bandwidth of the 100MHz bus speed. Potential problems with the Rambus technology are latency (which is, at the very least, no better than current memory implementations) and power requirements. In fact, many experts seem to agree that for uniprocessor, single user environments, Rambus technology will provide absolutely no benefits whatever! Despite this, consumers will be paying a very hefty price for the privilege of using it.
The one bright spot in this situation is that not all manufacturers have really bought into the scheme. Memory manufacturers must do so out of self defense, but motherboard and chipset manufacturers still have a choice. Most motherboard manufacturers believe that only very high-end systems will utilize Rambus technology initially, and that it will move to the desktop only after a few years. By then, they hope the costs will have come down substantially. On the other hand, chipset manufacturers have a big impact on Rambus. Without chipset support, even the best memory technology will wither and die. VIA currently does not plan to support Rambus technology, though they believe that eventually they will probably have to. In the short term, their focus is on DDR SDRAM, VCM and ESDRAM, and possibly SLDRAM. DDR SDRAM and SLDRAM offer the higher bandwidths that Rambus techology offers without the built-in costs, while VCM and ESDRAM have much lower latencies, thereby addressing a real bottleneck that exists today.
As with any technology, it is ultimately the end user who decides what succeeds and what does not. Several years ago, the Micro Channel architecture from IBM was considered superior to PCI, yet it failed because it was proprietary and PCI was not. Today, PCI is used in so many situations it is almost staggering. In my opinion, users need to voice their opinions to manufacturers in regards to performance and price for future memory offerings. If chipset and motherboard manufacturers believe that consumers will not pay the price for Rambus technology without knowing that they are truly getting some benefit from it, they will be hesitant to implement it. Intel and Rambus must also be pressured by the public to allow open discussion and let the technology win on it’s technical merits, and not just because of the marketing clout that Intel has.
In closing, I must say that I am not entirely opposed to Rambus, as there does seem to be some technical merit to it. Unfortunately, there is not enough public information available to make a truly informed decision, and those who know are restricted in what they can say. I recently asked for a point-by-point comparison between SLDRAM and Rambus from Micron, who is working on both. I received a very curt reply that essentially indicated that Micron would control this information and publish it if necessary. While I understand that this is required by their licensing agreement with Rambus, I do not feel it is in the best interests of the consumer or the vendor, or even the manufacturers. If those who are licensed for the technology (and therefore stand to make or lose money from it) are concerned about their inability to criticize it, those of us who will merely have to pay for it should be very concerned. This licensing is purely in the interests of Intel and Rambus, in my opinion, and as such I believe this technology should be seriously questioned and even opposed, if only on principle, until the current licensing practice is changed and users, vendors and manufacturers are able to fully evaluate the pros and cons of the technology before it lands in their servers and on their desktops.
Be the first to discuss this article!