NCSA’s Towns to appear on GPU computing panel
09.28.10 - Permalink
John Towns, head of the Persistent Infrastructure Directorate at NCSA, will be among the panelists during an Oct. 20 webcast on GPU Computing: Confronting the Need for Speed.
Along with Michael Wolfe (The Portland Group) and Mark R. Fernandez (Dell), Towns will discuss the benefits of GPU technologies as performance accelerators for computationally intensive and high-performance workloads and environments. Towns will highlight NCSA's experiences with Lincoln, a cluster than combines CPUs and GPUs to achieve peak performance of 47 teraflops.
After being deployed early in 2009, the Lincoln cluster was rapidly adopted by the scientific user community. Towns will highlight speed-ups and scientific results enabled by the system, including:
- GPUs provide more than 5 times speedup per node for Klaus Schulten's research team, which uses NAMD to study the organization and function of proteins and protein complexes within cells. Lincoln is two to three times as cost effective as a CPU cluster providing the same performance.
- Inanc Senocak, a Boise State researcher who studies atmospheric flows through complex terrain and urban environments, achieved more than 4.7 times speedup per node using the GIN3D fluid dynamics code. Lincoln is nearly 4 times as cost effective as a CPU cluster providing the same performance.
- University of Utah's Thomas Cheatham aims to develop better understanding of how proteins behave in solutions and how drugs interact with them. Using AMBER on GPUs, he has achieved more than four to 15 times speedup per node. Lincoln is three to 12 times as cost effective as a CPU cluster providing the same performance.
The webcast is sponsored by Scientific Computing and Dell. For details and to signup, go to http://www.ScientificComputing.com/GPUcomputing.