GPUs and the Future of Parallel Computing

This article discusses the capabilities of state-of-the art GPU-based high-throughput computing systems and considers the challenges to scaling single-chip parallel-computing systems, highlighting high-impact areas that the computing research community can address. NVIDIA Research is investigating an architecture for a heterogeneous high-performance computing system that seeks to address these challenges.
 
 

Authors: 
Stephen Keckler (NVIDIA)
David Glasco (NVIDIA)
Publication Date: 
Thursday, September 1, 2011