Carl, re: Hi wily; Re RDRAM manufacturing costs &c...
Made an infrequent visit to the SI thread today and came across this.
I assume by "RDRAM" you mean "RIMM", which includes the extra costs of the fancy PCB and the heat spreader. I've stated repeatedly that I expect the costs and prices of RIMMs to drop to around 50% over the costs and prices of SDRAM and DDR DIMMs. But man, I sure am waiting.
I seriously doubt you have any idea of the real costs of RIMM's over DIMM's. And you can't use resale price of products for comparison since in today's market DIMM's are selling below cost. With 128Mb equivalent chips at contract prices <$3.00, (and makers selling into the spot market at <$2.00), inventory levels at the makers in the 8 - 10 week range, this situation will last well into 2002. With the average PC shipped with 120MB's of memory, the migration to 256Mb devices will only make it worse.
Most of the high price now is due to the following facts:
(1) There are still relatively few suppliers of RDRAM, so there is a bit of a monopoly effect.
There is more than enough DRDRAM to meet the needs of the market. When 4i product becomes available in 2002 we'll see further decline in the learning curves. In the meantime, IMO, the price continues to be supported by Intel's subsidies both in the price of RIMM's and the back-end testing. I'm sure that Samsung is quite happy with the current situation.
The PC market itself is collapsing for reasons I've discussed before. IMO, it will take a couple of years of attrition among the makers, and wide acceptance of WindowsXP to resolve the industries problems. If XP doesn't become the dominant OS by 2003, matters will only get worse for the makers. There needs to be a compelling reason to ship PC's with >256MB of system memory if the industry is to recover. That's double the current average configuration.
(2) Unlike the case with Micron and DDR, no memory maker is pushing to make RDRAM mainstream. This is because they know it can't be done. RDRAM has higher costs that add up to around a 50% premium, and this is too high to allow RDRAM to become mainstream.
Since your definition of "Mainstream" is >50% of the total DRAM market, neither DDR or DRDRAM will meet that criteria in 2002. Within Intel's Mainstream segments, it's obvious that P4/DRDRAM will continue to grow in the high-end and mid-range segments. Consider that MSFT, Intel and the PC industry is reported to have committed $1 billion to marketing WindowsXP and the best solutions for that OS. Guess what MSFT, Intel and DELL will be pushing. Never underestimate the power of marketing.
(3) The memory makers are pretty sure that 4Q01 is going to be the peak quarter for RDRAM production and from there it will decline. The R&D for RDRAM was expensive, and so is the cost to convert an SDRAM line over to RDRAM. Since the RDRAM market will contract after the end of the year, those lines will only be able to
produce RDRAM for relatively short periods of time. Most memory lines are run for about 3 years, RDRAM lines are being set up to run for only about 1 year. Anyway, the short run time means that the R&D and setup costs have to be amortized over a much shorter period than with DDR or SDRAM.
Pure speculation on your part.
Regarding (2) and Micron's driving DDR prices down, it is probably important to note that Micron's DDR is the most standard, in terms of being compatible with the most motherboards, and also the most stable and overclockable, according to the benchmark sites. (In fact, I've never seen a report of Micron DDR being incompatible with anything, while there have been plenty of reports of other makers' DDR having problems.) From this, I would say that Micron's strategy is to make themselves the DDR type that all other types are compared against. It will likely sell for a premium (over other DDR types) by the end of the year.
Did you read and understand what you said? Only one supplier of compatible/stable product?
The three memory types will co-exist. How they split market share will be determined by price/performance and marketing. If you think that Intel intends to abandon DRDRAM at some future date you're going to be proved wrong.
Here's an excerpt from an e-mail I sent to Blake's Rambusop's members.
"-SDRAM
A real indicator of the ill's in the PC industry. After listening to the Micron and Infineon CC's, I come away with the following. Inventory at the makers is 8-10 weeks. While bits shipped will be flat sequentially, Q2 revenues will decline 30-50% with no signs of improvement for many quarters. Contract pricing for 128Mb equivalents is below $3.00, and shipments into the spot market are increasing. There is speculation that spot market purchasers are betting on some recovery and are buying chips below $2.00. As I recall, Infineon stated that shipments into the spot market have grown to about 20% of total shipments. The makers seem to be committed to maintaining capacity at 100%. It's cheaper to keep fab's running at capacity, even if it's at a loss. According to Micron the average PC now ships with 120MB of DRAM. The problem is that the industry is moving to 256Mb DRAM and there needs to be significant increase in MB's/PC to resolve the capacity issues. Even with return to normal growth in PC shipments I think we have a capacity problem. Again, while the industry is hoping that WindowsXP will create that demand, I wonder if it will happen at a rate sufficient to resolve the capacity issues in the next year or so. The real issue is that the migration to higher densities on larger wafers with tighter design rules creates significant increases in bits out. Without demand to consume this product, inventory builds and prices will remain in free-fall. Indications are that 2002 will not create sufficient demand to bring supply/demand into equilibrium. We may see some attrition in the weaker suppliers."
I have no doubt that 128Mb equivalent RDRAM will sell well below $5.00 in 2002. If so, the price/performance at the system level will have more impact on purchase decisions than the relatively small price differential of the memory types.
(And I still think SDRAM will have major share of the DRAM market through 2002).
JMHO's
PS. I still intend to stay away from this thread, (lurk once in a while), but I'll respond to your spin. <VBG> |
