I don't disagree with you about the questions for RDRAM. Obviously the market is re-evaluating TER due to the RDRAM ramp up, and that is good for me. As for the die size issues, where does your data come from? Mine comes from the MU conference calls. In them, MU stated that most people may see a smaller die size penalty like 10-15%, but that those who were already very efficient in the use of silicon real-estate (like MU) would see a larger penalty, more like 25%. They further stated that they expected that including packaging, RDRAM costs would be 50% higher, not the 30% you cite. I'm not saying you're wrong, just asking what your source is. Is it coming from other DRAM makers?
As for who will pay the penalty, remember that DELL was the first to implement SDRAM, and sold a lot of computers at a higher price than the competition based on having SDRAM in all computers (even though SDRAM provided little measurable benefit in computers with burst cache). With a 50% penalty, 128MB of SDRAM will cost maybe $75 more, and with a 30% penalty the difference will be more like $50. Who will ultimately decide if RDRAM is successful? The market! If consumers and businesses are willing to pay $100-200 more for a computer with RDRAM, RDRAM will be a success. Will they? I don't know.
I am pretty sure that when you factor in that cache hit rates are over 90%, the difference in performance of an RDRAM system over an SDRAM system will be barely measurable, if at all, for most applications. But that was true of SDRAM as well, yet consumers demanded SDRAM. So we will just have to see how it plays out. My guess is that RDRAM will be used in some high end workstation systems and on a few servers, but not the lower end systems. But it's only a guess.
Thanks for your informative posts,
Carl |
