As inference proliferates to edge servers and endpoints, memory solutions must balance performance, cost, and power ...