Hi The Propher; Re: "In fact, they will sell RDRAM for cheaper than SDRAM, just as SDRAM now sells for cheaper than EDO."
Before you believe the lies spoken by Rambus management with regard to pricing of RDRAM vs SDRAM, you should recall some of their previous statements on this subject. This one, in particular, is a hooter:
Memory-bus wrangle September 23, 1996
"Rambus expects to see production of 64-Mbit RDRAMs [Rambus-specific DRAM] by the end of 1997," said Subodh Toprani, the company's vice president of marketing. By that time, price should be at parity with SDRAMs.
techweb.com
Now which part of "1997" are we living in? LOL!!! Heck, the trend in RDRAM/SDRAM prices is flat at 5 to 1 for the last two months. When is all this 1 to 1 priced RDRAM going to show up?
Those of us who live in the memory industry know that SDRAM was brought below DRAM in price by the memory makers, as an aid to the conversion of the industry over from DRAM to SDRAM. A few memory makers did this by guaranteeing prices for SDRAM at parity to the price of DRAM. Right now, as mentioned in my post here: #reply-15329994, Micron is guaranteeing DDR266 prices no higher than PC133 in 2Q01. This is an exact historical repetition of the DRAM to SDRAM conversion.
On the other hand, RDRAM has been in production for a year and a half already and no memory maker has stepped up to the plate with a similar guarantee. Instead, the memory makers have repeatedly stated that RDRAM is more expensive to make and requires a larger die than SDRAM. They've suggested that they will reduce the cost of manufacturing RDRAM, and hey, who wouldn't. But they have not even hinted that they are going to start guaranteeing large volumes of RDRAM at SDRAM pricing. Even Samsung, when they said that they would reduce the die penalty for RDRAM did not say that they were going to sell it at the price of SDRAM, or that the cost of producing RDRAM would drop below that of SDRAM.
RDRAM could only get cheaper than SDRAM if RDRAM were the next standard memory, but that is not happening. Instead, new designs are going to DDR. Everyone in the industry knows this, and probably most of the people on Wall Street know that Rambus is only valuable to the extent that they collect royalties on SDRAM and DDR.
Right now, this distinction, that Rambus' fortunes are riding on the patent lawsuits is of no big concern, as the patent lawsuits are still to be decided. But if Rambus loses their royalties on SDRAM and/or DDR, it would be a good idea to have a realistic understanding of where the memory industry is, where it has been, and where it is going. These issues are not digestible to simple one line slogans that morons with 9th grade educations and no experience in the industry can understand.
Micron, in addition to publicly guaranteeing to sell DDR200 at PC133 prices, did so, enough to make the stuff widely available on pricewatch, and this dropped the price for it to under half within a few weeks. As of today, 256MB DDR DIMMs are available from 36 sources on price watch at prices cheaper than the cheapest 256MB RDRAM RIMM on the same source. Mainstream memory is all about money. Hey, the story is over, DDR is the next memory standard.
To remind people of how SDRAM was brought into mainstream use, the following articles should suffice:
Fujitsu Gets In Synch June 27, 1994
Fujitsu Microelectronics Inc. said last week that it plans to stop selling conventional DRAMs and will instead concentrate its resources on the development of faster SDRAM
...
The 20% to 30% price difference between SDRAMs and DRAMs means that the initial markets for SDRAMs will be in high-end computers such as mainframes and workstations, company officials said.
techweb.com
Note that Fujitsu was a major memory producer at the time, and they announced that they were quitting production of DRAM and going entirely to SDRAM. Converting a factory from SDRAM to DDR is essentially free, compared to the DRAM to SDRAM or SDRAM to RDRAM conversion. Consequently, Micron doesn't have to announce that they are converting entirely to DDR three months in advance. They can build die that can go either way, and produce enough to satisfy the DDR market up to their full production without more than about a month's notice.
Samsung grabs for 64-Mbit DRAM lead December 11, 1995:
In a highly unusual move, Samsung is publishing its pricing for the new parts, hoping to persuade customers that now is the time to jump to 64 Mbits.
...
"Right now, SDRAMs are selling at about a 25-percent premium over similarly organized EDO parts. In the first half of 1996, we will drop that figure to 16 percent, then 8 percent in the second half. In the first half of 1997, there will be no premium for SDRAM."
techweb.com
The way that the memory makers influence engineers (like me) to choose between memory technologies is by guaranteeing my procurement department that the new fangled parts will be available, and available at no premium in price. With SDRAM, that was done by Fujitsu and Samsung (among others). With DDR, it is now being done by Micron. Micron says that they are now shipping more bits than anyone else in the world, so they can make us engineers choose what they like. I wrote a more complete description of the DRAM to SDRAM changeover back in May of last year, if you're interested further in the subject: #reply-13764437
Since Micron produces about 20% of the world's DRAM bits, and they can easily switch between SDRAM and DDR production, this means that Micron can arrange for world DDR production to be as high as 20%, even ignoring production from Infineon, Hyundai, Samsung &c. This means that Micron can guarantee to engineers (like me) that they can produce enough DDR to satisfy my build requirements as long as DDR doesn't use up more than 20% of world production. DDR is supposed to reach something like 19% in 2H01, according to slides at the recent DDR show, which are likely overly optimistic. This means that I'm about as safe as I can get specifying DDR in a product. Micron will have the parts. There won't be such a long line of companies ahead of me that I get put on allocation. The parts won't cost more than SDRAM. Under this scenario, of course I'm going to spec DDR.
With RDRAM, by comparison, there are no guarantees of when parity pricing will, be it at the end of this year or any other time. There are only suggestions from parties (like Rambus) who have been horribly wrong in the past. It isn't easy for the memory house to make the stuff, so if demand for P4s is above what was predicted I'm likely to be put on "allocation" (unable to buy it). The upshot: An engineer would have to be a total moron to design RDRAM into a product (in 2001) that could use DDR instead. There was a time when RDRAM seemed like an idea that had a chance, but it all went away with Camino, and the collapse of (real) Intel support. That Intel would pull its support was obvious to us all in the fall of 1999, but it probably did not become obvious to most people until Intel announced SDRAM and DDR support for the P4. Back in '99, you had an excuse for not seeing that RDRAM had failed, to see it then required insight into the relationship between the memory houses and the memory designers (and buyers). Now you don't have an excuse, it should be obvious to all.
-- Carl |
