Place: 4274 Chamberlin Hall (refreshments will be served)
Speaker: Jing Li, Department of Computer Sciences
Abstract: The confluence of disruptive technologies beyond CMOS and "Big Data" workloads calls for a fundamental paradigm shift from homogenous compute-centric system to heterogeneous data-centric system for better innovation, competition and productivity. With the objective of rethinking data-centric system from ground up, through a concrete example, I will show how to leverage emerging memory technology such as phase-change memory (PCM) to realize a new IC building block for future data-centric system. A novel chip was designed and fabricated for the first time, blurring the boundary between computation and storage, i.e., it can either be configured as a compute unit - a high performance search engine or as a storage media - storage class memory. It achieves >10x area reduction compared to homogenous CMOS-based design at the same technology node and reliably operates at ultra-low voltage down to 750mV. In the talk I will briefly highlight a few critical enabling techniques from material, circuit, architecture and algorithm perspectives. I will also highlight the major research activities in my lab in developing collaborative software/hardware solutions to address classical von Neumann bottlenecks.