Ok, went back and listened to the whole replay. It's not all that easy trying to understand what's being said, but I think I've cleared up a couple of things.
First EcoRAM will do 10k writes not 1M as I reported earlier. I don't know how load leveling works, but as EcoRAM nears its 10 year life span the amount of useable memory must be a small fraction of the original 32GB. So, given the advantages of having as much memory as possible I'm not sure just how long the practical life of an EcoRAM stick will be for users.
Second, in the type of searching server application envisioned the data base will be loaded from somewhere else via the network so writing, particularly random updating, isn't going to be a problem. However, it's quite apparent that some significant progress must have been made to the write process as a 1TB load which would take 5 hrs. using NOR will take just 30min using EcoRAM. Raj said that they can load at network system speeds all day and never have problems. Virident probably had a hand in the improved write ability.
Third as Raj said once you have the persistency of memory all sorts of things become possible that have never even been thought about before. Where all this is going is anyones guess, but the possibilities seem staggering. It's very likely that while EcoRAM is going to play a part in getting to the promised land of retained memory it probably wont be the final solution. Still, it could be very profitable for as long as it lasts.
In any case, Berni and Raj seem hell bent on getting this first application out of the door by the end of the year. Where that leads next will probably depend on the innate capabilities of EcoRAM and what new capabilities SPSN and Virident can coax out.
I wonder how NUMONYX is coming with PCM? I can't think of anything else that could challenge EcoRAM currently. |
