It would appear that after a long and somewhat painful journey, DRDRAM is all but dead. One reader recently sent in the following email: “I ordered a Dell Dimension XPS 733mhz with 400mhz RDRAM on November 30th and they are having trouble shipping because they say there is a problem with getting the ram chips….They first gave me a delivery date of Dec 15th, now it’s after Jan 20th.”
The first real clue about the health of DRDRAM was given on Dec. 9 when Rambus announced that they would be pursuing an “increased focus on providing its technology to the communications market and for other chip-to-chip applications.” I took this announcement to mean that the company was no longer convinced that the PC market would provide the revenues they had originally anticipated.
Several weeks later, the other shoe dropped when it was reported that Intel had called for an ‘affiliation’ with the 6 major memory manufacturers (who together own approx. 80% of the DRAM market) to ‘define the next generation memory standard‘. Initial reports indicated that this was a move by Intel to try and define a new proprietary standard, however calls to sources within the memory industry indicated that this is not the case.
It was mentioned to me that on Jan 10, an agreement between Intel and Rambus expires which would have awarded certain considerations to Intel if DRDRAM had gained a large enough market share (reported to be between 10% and 20%). Though I am obviously unable to confirm whether this is true, the timing seems to coincide with these other events. The consideration was reported to be that the rights to the technology would transfer to Intel if this market share had been reached. Since this did not happen, Intel seems to have actually been burned twice – first for spending so much time and effort to develop and market the failed DRDRAM solutions, and second for essentially losing control of the technology itself.
The fact is that DRDRAM was not Intel’s first foray into ‘manipulating’ the memory industry. In fact, Intel essentially decided that SDRAM would be the next standard after EDO, shunning the BEDO solution that Micron had already developed. The reason for this is not simply that Intel wants to control the memory market, they actually need to know in advance what the memory standard will be so that they can incorporate it into their chipset and processor designs. This is one reason that Intel chipsets have traditionally been better performers with regard to memory transfers.
The ‘affiliation’ that Intel suggested seems to actually be a white flag, conceding that perhaps the memory manufacturers and designers are in a better position to determine what makes sense in the market place than Intel is. My sources indicated that no mention of proprietary technology was made, simply that Intel wanted to create an ‘independent’ group that would define memory standards far in advance of their implementation. Of course, that group didn’t include rival chipset maker VIA, but that is to be expected in this recently competitive arena.
DDR and DDR II
When Intel first capitulated and gave support to 133MHz SDRAM, they called it a ‘transition’ product between SDRAM and DRDRAM. It is safe to say that PC133 can still be called a transition product, but the transition is to DDR. Both VIA and AMD have announced their intention to include DDR support in upcoming chipsets. Micron has already demonstrated systems using their Samurai/Shark chipset using DDR SDRAM. They have also provided the design to other chipset manufacturers, including VIA.
Virtually all sources within the memory industry that I have spoken with have indicated that they believe DDR will be the dominant memory by the end of 2000, transitioning to DDR II sometime in 2001. The exciting thing about DDR II is that it will apparently be a packet based memory design, using some of the technology developed for SLDRAM.
Since November, memory prices have been slowly floating downward. Though not anywhere near their all-time lows from last May/June, spot prices appear to have settled down to about $8 per chip (64Mb parts). Contract prices remain stable at about $9 per chip, and will likely remain there for some time. Memory manufacturers seemingly have learned to control production and forecast better than in the past. In addition, manufacturers continue to leave the commodity DRAM market in order to focus on the more lucrative specialty DRAM market.
Last year LGS and Hyundai merged, making them the largest memory manufacturer in the world, followed by Samsung and Micron. The remaining three manufacturers with significant market share are NEC, Hitachi and Infineon (formerly Siemens). Between these six manufacturers, they own approximately 80% of the global DRAM market. There are some rumors that NEC and Hitachi may merge, making them the #2 or #3 manufacturer in the world. Taiwanese manufacturers (excluding outsourced manufacturing from Japanese and Korean firms) own about 5% of the market.
Odds and Ends…
I must admit that I don’t keep very close tabs on the storage market (i.e., hard drives, etc.) but it was recently brought to my attention that the cost per megabyte of storage continues to decline at a dramatic rate, and is now approaching 50 cents/megabyte! (Correction: make that .5 cents/MB!). 30GB+ drives are now the ‘sweet spot’, with several models being offered well below $300. I am fully expecting to see very affordable 100GB drives available by the end of the year.
It would seem to me that we are approaching a point where capacity is not the biggest concern. With processors and memory speeds increasing dramatically, the next major performance bottleneck is the I/O subsystem. Expect to see some major developments in this area in the coming year from both Intel and AMD.
As always, if you have reliable information, or see information published elsewhere that might be useful, please let us know by emailing firstname.lastname@example.org.