
Nvidia is reportedly shifting its Rubin Ultra AI accelerator to an eight-high high-bandwidth memory (HBM) configuration from a previous 12-high design due to rising memory costs, focusing on bandwidth efficiency and overall system economics.
News in Brief

Nvidia is reportedly shifting its Rubin Ultra AI accelerator to an eight-high high-bandwidth memory (HBM) configuration from a previous 12-high design due to rising memory costs, focusing on bandwidth efficiency and overall system economics.