Ultralow-Power Implantable Electronics
Seetharam Narasimhan, Swarup Bhunia · 2013
Large-scale high-performance computing (HPC) systems have historically been designed and deployed to assist the scientific community with discovery. These systems have evolved from strictly proprietary monolithic designs to include commodity-like networks-of-workstations (NOWs) called “clusters.” Clusters outwardly resemblemachines found in traditional information technology (IT) data centers; however, HPC data centers tend to be extremely dense, and almost always incorporate state-of-the-art interconnection and storage technologies to counteract speed-of-light latencies associated with data movement. Therefore, HPC systems are generally more complex than commercial IT systems, requiring new mechanisms to handle increased reliability risk. These factors, high density and increased reliability risk, invariably translate into increased power requirements, presenting new challenges and opportunities in the areas of power and thermal management.