Preparing for Exascale: Efficiency from the Siliscon to the Data Center
Michael K. Patterson
Exascale Systems Architecture Pathfinding Group
Intel Corporation
Dupont, Washington
Abstract
The trend towards more and more powerful supercomputing has continued unabated for the last 20 years. This trend has been based on Moore’s Law and the skill of the HPC system designer, but also on the lack of significant external constraints. Exascale is different. A goal of 20 MW has added a significant new constraint that changes the landscape. In this talk we will discuss why Exascale is important, how it might look different, and what we need to do now to prepare for it.