Seminars

A Perspective on GPU Computing

Tuesday, May 1, 2012

Dr. Michael Shebanow

Abstract

GPUs have fundamentally changed the playing field of high performance computing. Starting out as devices intended only for the display of 3D images, GPUs are now used as supercomputers – attached processors used to accelerate computationally intensive applications. In this talk, using NVIDA GPUs as a basis, I’ll provide a brief history of the GPU, the evolution of GPUs into computing devices, understanding their performance characteristics, and the challenges that lie in attaining high performance from these devices.

Speaker Biography

Michael Shebanow joined NVIDIA in 2003. While at NVIDIA, he has worked on the Tesla product family (G80, GeForce 68xx series) and was one of the architects of the Fermi (GF100) family, managing the shader processor architecture team (covered 5 blocks including the SM & L1). After Fermi, he worked in the research group investigating next generation graphics and unified programming models for GPUs. Currently, he is an architect working on the Denver ARM core project. Prior to NVIDIA, he has managed the development of a number of processors in multiple architecture families (x86-32, x86-64, SPARC v9, 68k, m88k), and was a representative representing Motorola in the Power PC architecture definition committee. While a graduate student at UC Berkeley, he was one of the original members of the HPS research team (superscalar, dynamically scheduled processor architectures) (started 1984). Dr. Shebanow holds 30 patents in graphics, processor design, and disk controller areas.