Re: I believe that drdram actually has a cost penalty of about 40%
It may be the drdram will always cost enough more than other technologies to impact its acceptability. Since drdram has to include a mini-chipset in the module to control the 128 100MHZ dram channels that it buffers to send down its 600 to 800MHZ bus, there's got to be a chunk of 400MHZ logic in that controller, together with supporting circuitry. SDRAMS have relatively less logic, hence smaller die size and simpler design - and of course on top of whatever manufacturing differential exists, there are the fees paid to RMBS.
The companies with RAM foundries, with their huge capital investments at risk, can't be too happy about paying what will amount to something between a chunk of and all of their profits to rambus which has almost no capital at risk, just gain at the expense of the memory fab companies. It'd be pretty galling. If they can make money with rambus, they most certainly will - but I would expect a tendency to let rambus founder from the companies who must produce it. I think that in the long run, this may be the most important factor. DDRDRAM and high speed DRAM technology was selected by, and invested in internally by, the RAM industry. Intel has had to pay hundreds of millions of dollars to get them to implement rambus (200 to 400 million to micron alone).
Micron, Samsung, Hitachi, etc. have all spent their own hundreds of millions to develop and produce 166MHZ SDRAM and 266 DDRDRAM for the same time frame as 800MHZ DRDRAM. Can they influence the market? Or are they totally at the mercy of intel's decision to push rambus? I don't know, but I've seen the not-invented-here effect in other places, and it can be very strong. |
