To: JD_Canuck who wrote (87328 ) 12/6/2003 10:45:23 PM From: Bilow Read Replies (2) | Respond to of 93625 Hi JD_Canuck; Re: "The articles i saw seemed to indicate about a 15% lead for RDRAM, have you seen any others? The surprise to me was the fact that the industry doesn't seem to care for the performance more efficient memory provides anymore ... " Okay, assume RDRAM is 15% faster, though this depends on exactly what you're comparing and is rather difficult to quantify. The problem with RDRAM is that it is far more than 15% more expensive than DDR. The cheapest PC800 512MB stick I could find on the same source was $160. That's 55% more expensive for half the memory. A 512MB stick of PC2100 DDR is only $53. Even a 1GB stick is only $105 (all priced at pricewatch). That means that RDRAM is about 201% more expensive than DDR. The whole problem with RDRAM was that it was too expensive. I told you guys years ago that RDRAM was never going to be the mainstream memory, and that it would always be too expensive. I was right, and the reasons I gave then are still true now, and will still be true later when Rambus tries to convince the world to convert to whatever weird memory type they come up with next. The basic problem is that Rambus' memory interfaces are extraordinarily expensive and difficult to manufacture. This is true both of the memory chips and the boards that use them. If these difficulties were only true of the chips, and RDRAM motherboards were easy to make, Intel would have succeeded in converting the world over to Rambus, as Intel carries a LOT of weight in the memory industry. But since RDRAM boards were much more difficult to make than DDR boards, this forced a LOT of smaller memory users to choose DDR instead of RDRAM. That meant that no matter WHAT Intel did, there would always be at least a small DDR market place. And since DDR chips are considerably cheaper to produce than RDRAM chips, that meant that it would be impossible for RDRAM to obtain enough decrease in price due to production size to get as cheap as DDR. And it is a very simple fact of the memory industry that when choosing between two memories of vastly different prices, the big users ALWAYS choose the cheaper memory. Because of these very simple facts about the memory market, Rambus probably never had a chance. But when the i820 blew up in late 1999, that small chance was obviously dead, dead, dead. With Intel screwed up by Rambus, the rest of the memory industry concluded (correctly) that Intel would dump Rambus and that was the end of that. Re: "... and has lulled us all into accepting a second rate standard. " (a) The memory designers who chose DDR over RDRAM knew exactly what the limitations and advantages of each technology was. The memory designers were not lulled into anything. I told you that we were rejecting RDRAM way back in 1999. It was the mom and pop stock buyers, people who had never even laid out a printed circuit board, were the ones who were "lulled" into believing that RDRAM was goign to win. (b) And like I've said over and over on this thread, mom and pop aren't the ones that get to "accept" what memory type will be standard. It's memory designers that choose the memory. Mom and pop stand in line and buy whatever comes out of the factory, unless they're Rambus believers who will only buy RDRAM, in which case they're now up sh:t creek, LOL. All that Rambus BS about how much better RDRAM was than DDR was just BS. It was never true. RDRAM is only, as you yourself admit, at most 15% faster than DDR. And that matters only in very specialized situations. For the typical memory use, DDR is equally as fast as RDRAM and a hell of a lot cheaper. The guys who speced RDRAM into their systems mostly did it in a context of Intel pushing the technology onto the world, and when Intel backtracked, their designs had gone too far forward to replace the memory system, so they were forced to ship with RDRAM. And there are always going to be a few designs where RDRAM is more efficient (from a cost / performance point of view) than DDR. The essence of the problem is that for the bulk of the memory industry, the performance / cost metric for DRAM is not [bandwidth / cost] (which RDRAM wasn't that good at anyway), but instead [memory size / cost]. That means that they want cheap bits, not fast bits. -- Carl