Presumably for best performance this will need to be accessed directly via QPI/DMI rather than via PCIe or similar? - you'd be throwing away the benefits going over a slower bus such as NVMe
I'm reasonably sure this'll need a new CPU generation to support it, is that a good assumption?
What CPU generation will this work with, when will it be released and do we know if it will it be introduced on the consumer (i3/5/7) range or the professional (Xeon) range first?
When they say '1000 x faster' do we know if that is that across the board, i.e. is latency reduced equally?
Do we know what sizes these initial units will come in? They mention 128GB blocks but will they be shipping JUST 128GB devices or 'multiples'?
Thanks a ever TTL!