Driving Innovations from Mobile to Exascale using 3D Memory
|Event Date:||March 11, 2015|
|Speaker Affiliation:||Georgia Institute of Technology|
|Contact Name:||Professor Sam Midkiff
Computing systems face two daunting challenges: slowdown of Moore’s Law and memory bandwidth bottleneck. 3D stacking is a game-changing technology that can simultaneously address the two challenges by vertically integrating multiple dies within the same chip. It extends Moore's Law for another decade or more and provides large amounts of high-bandwidth energy-efficient memory storage. However, incorporating 3D memory into computing environments requires innovations across the layers of the system stack from hardware and architecture to algorithms and software.
In this talk, I will describe why 3D stacking is happening -- especially in memory -- and how we could exploit 3D stacked memory for next-generation computing. Specifically, I will discuss how to practically architect gigabytes of stacked memory for two primary use cases, large caches and system-visible memory. I will then share my vision that 3D memory leads to a new paradigm of data-centric computing and enables near-data processing -- the notion of moving computation to data.
Jaewoong Sim is a PhD candidate in the School of Electrical and Computer Engineering at Georgia Institute of Technology. His research interests are in the interactions between microarchitecture, compilers, and operating systems to efficiently enable emerging architectures and technologies. His dissertation work focuses on addressing the challenges of enabling 3D die-stacked memory, including scalability, resiliency, and programmability. His work has earned him several awards including IEEE MICRO Top Picks and Outstanding Graduate Research Assistant Award from Georgia Tech, and has resulted in eight patent applications with AMD and Intel. He received his BS from Seoul National University in 2007.