That will depend on both the nature of the processing load and the computational primitives implemented in the "memory units".
Basically, though, this sounds like [very roughly speaking - you lose a lot of knobs and whistles] turning machines such as the raspberry pi into integrated circuits and giving them a memory bus interface to the outside world. Put differently, it's [again, very roughly speaking] the cloud model for computing implemented at the hardware level. I'm curious how the powerup / initialization process will work here. You need some minimum level of functionality and an initial state that does not crash before the rest of the system can work. This has all sorts of implications (mostly in terms of new failure modes when done wrong). Thanks, -- Raul On Thu, Oct 26, 2017 at 2:00 PM, 'Skip Cave' via Chat <[email protected]> wrote: > IBM scientists say radical new ‘in-memory’ computing architecture will > speed up computers by 200 times > > https://goo.gl/cA27v4 > > Would this new memory architecture make parallel processing more efficient? > > > Skip Cave > Cave Consulting LLC > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
