Steve, Re: "That's my point exactly. RDRAM may be expensive for maybe 2 weeks, and then
when no-one is willing to pay the premium the prices will come down and thr
profit will go out of it. Face it DRAM is in the toilet for a long time..."
Don't you think that RDRAM is targeted for top of the line servers based on Xeon, and then Merced (plus Alpha, SPARC?) CPU chips? Of course, it's overkill for desktop PCs, whether of standard issue business or home variety. With what I know about the development history of Merced, and that of Rambus RDRAM, it sounds logical that the former begat the latter. Intel had the foresight to realize a bottleneck in the main memory bandwidth area, and the wherewithal to do something about it by getting alongside Rambus. If the market for very powerful NT servers (and don't forget Solaris on Merced) is as big as it's being cracked up to be, Rambus and RDRAM should go along for a nice ride.
Not you Steve (I don't think), but someone said 'why do you need very fast main memory, what with the sizes of L2 cache that are being offered.' The answer is, if you have enough very fast processors, and we're talking up to eight in SMP configurations just for Xeon, you can never get too much, or too fast memory.
Tony |
