|> I think I read that one. It was right before dram bandwidth took a |> sharp upward turn from something like 400mhz to I think |1600mhz today. |> Prior to that, it exhibited less than half the rate of change, |> increasing from 100 mhz in 1998 to 400 mhz in 2005. |> | |ah, then... perhaps tony can weigh in with a: "Yea, all good now" or |"well, it's not quite as simple as that... + explanation?" ?? I was |under the impression that DRAM wasn't the issue, but I'm fuzzy now as |to the details.
DRAM is a clear case in point where our growth exceeds what technology has delivered. Thus we simply can't "throw hardware at the problem". Now I have to confess that I'm certainly not the world's expert on DRAM technology, but nothing that I've seen so far truly changes the equation. First, recall that the focus is on commodity DRAM, as that is the segment that drives volumes and thus prices. Yes, we can always go off of the commodity cost curve, but if we're willing to do that we can also just throw gobs of SRAM at the problem. Again the point is what we will be able to do for constant cost. So far, what I'm seeing is that faster memories are not going to be the mainstream. Yes, that may change, but we have yet to see how that is going to play out. >From the technology front, my understanding is that these higher bandwidths are being achieved through more channels (i.e., replication) and by migrating from DDR2 to DDR3. This improves the bandwidth to memory, but does not improve the latency of the underlying memory cells. (Ref: http://en.wikipedia.org/wiki/DDR3_SDRAM). Unfortunately, BGP has an annoying tendency to do fairly random access (tree walks) through large data sets and then perform read-modify-write operations (e.g., adding a path to an existing route and then performing the best-path computation), certain operations such as caching and pre-fetching have limited benefits and latency remains the key concern. In short, unless there's a significant change in the underlying technology of the single DRAM cell, things aren't going to change significantly. Regards, Tony _______________________________________________ rrg mailing list [email protected] https://www.irtf.org/mailman/listinfo/rrg
