Another Rambus/DDR article:
IBM, Reliance lead march toward Direct Rambus alternative
the companies foresee using PC133-standard 133-MHz SDRAMs first, and then double-data-rate (DDR) SDRAMs.
...
a number of Taiwanese chip-set companies said they will exploit the 133-MHz SDRAM architecture to serve low-cost systems.
techweb.com
The project I'm on is a little short of DDR chips... I hear that Rambus chips are also quite scarce, from an interviewing Intel(!) engineer. Both types are being reserved for customers who have a hope in hades of eventually ramping up production to large volumes, so small guys like me get left out. (All is not lost, I've got 4 coming in, and that is enough to get some pretty good testing going.) When everybody knows there is going to be a shortage, everybody tries to get there orders in early, thereby causing a shortage. (Just like Johnny Carson and TP.)
Incidentally, the article notes the problem that the memory makers have guaranteeing hold times, given short access times. When I first found out that the DDRs from Samsung had an access time of 6ns and a hold time of 3ns, I was stunned. I still find it hard to believe that they can spec a chip that tightly over temperature, voltage and process. The ratio of fastest to slowest propagation delays for ICs has been expanding for years, as processes shrink. I guess the memory makers must be going to some sort of compensation circuit. As far as I am concerned, they ought to blow off the hold time, and give the fastest access times possible. It is generally a lot harder to take advantage of long hold times, than short access times...
-- Carl |
