Hi jim kelley; Re the link on (yech!) yahoo. The guy says a lot of things I agree with completely. He writes:
I have no specific information to answer your question. I do have some significant number of years of experience in the chip business and can tell you what I think will happen.
Usually it takes about two years to go from concept to first samples of working silicon for non-trivial custom logic/asics for advanced applications. Some take longer (like a new CPU project) some take less time (like a microcontroller).
A complex ASIC needed to attack a leading-edge networking application would need a significant effort on both the chip design side and the end-product design side most likely: if it is a killer product, it will be a complex design.
If one company is doing both sides of the design, you'll see the end product sooner than if a chip company is doing the chip design and a box company does the system design/box design: design communications work better under one roof than when the design groups are located halfway around the world from each other.
Rambus have announced a number of communications related deals over the last year or so. Some of these deals are with companies that make chips for open market sales, and some are with companies that make end products.
Rambus should begin to see some system level communications product introductions within the next year or possibly sooner applying the timeline I described above. It is anyone's guess how quickly the revenue will ramp, and that is dependent more on the execution of Rambus' licensees than it is on Rambus' execution.
messages.yahoo.com
This is exactly why I wrote back in November that even though it was evident that RDRAM was dead, dead, dead, you would still see new products announced with RDRAM for most of the next (i.e. this) year. Though it takes something like 2 years to bring a product to market, you can redirect the memory type up to about a year or so before you ship it customers.
After that moment, changing the memory type will cause your schedule to slip. That is why we are still seeing RDRAM design wins. For the same reason, the vast majority of DDR design wins will not be evident until 4Q00 to 1Q01. It just hasn't been long enough that RDRAM has been dead for the industry to switch over to DDR.
To reiterate, the way that industry dumps a technology is as so: (1) The design engineers dump it. (2) Twelve months later, new products are no longer announced with it. (3) Six months or a year later, shipments of the technology quit increasing and begin a slow decline. (4) Another 2 or 3 years, the product finally dies away.
RDRAM is now between stages (1) and (2). EDO and FPM are in stage (4). SDRAM is in stage (3). DDR is still in its infancy.
The nichedom of RDRAM is now quite fully established, and this means that prices should eventually stabilize at niche pricing, around 50% over SDRAM. That doesn't mean that prices won't go above or below that, just that the long term average is going to be around there. (RDRAM still has a ways to drop before it gets to 50%, the Register article over estimated the price of SDRAM.)
You should note that a lot of current networking DRAM problems are getting solved with huge embedded memories. Extreme high bandwidths per pin is what RDRAM is for, but DDR is coming up fast on Rambus like frequencies, and at a much lower cost.
By the way, I'm still waiting for MAJC to show up, it's getting kind of boring. It was supposed to be out in June, but no go.
-- Carl
P.S. Maybe Yahoo ain't so bad... |
