<< Explain how latency increases with faster clock speeds on SDRAM. The only reason latency would increase is if you used memory that's rated for say, CL2 at 266 but that same module only handles CL3 at 333. That is not the same as the entire architecture automatically having higher latency at higher speeds (whereas RDRAM's architecture does get lower latency at higher speed by design), it's simply that the modules available right now are not rated for the same latencies at higher speeds. Once DDR333 becomes more of a standard and common item, CL2.5 or CL2 will be just as common as it is with DDR266 now. >>
I'm not referring to column address strobe latency, but rather, access latency (I don't know the technical term for it, so bear with me). Since I'm not an EE, I'll explain this the easiest way I can: examples. PC2100 DDR SDRAM has a latency of 200ns (remember not the latency of the memory chips themselves or the CL), whereas PC800 RDRAM has 300ns (hence SDRAM's advantage in office apps). PC1066 RDRAM has a latency of 230ns, and I believe PC2700 has 230ns as well. Now, I could be grossly mistaken here, in which case please correct me. Assuming even that I'm incorrect, the higher CAS latencies at higher speeds of DDR SDRAM are undesirable compared to the lower latencies of RDRAM. It has also become apparent that DDR SDRAM is reaching its limits, whereas RDRAM is not (though this point may be erroneous for this thread).