Communications of the ACM
Ncsa Installing 153 Teraflop Supercomputer
NCSA's Lincoln hybrid supercomputer will be replaced by Forge, which will combine Dell servers utilizing AMD processors and M2070 Nvidia Fermi graphics processing units.
Credit: ACM Queue
The U.S. National Center for Supercomputing Applications (NCSA) has begun installing Forge, a 153-teraflop hybrid supercomputer that contains both central processing units and general-purpose graphics processing units (GPUs). It will replace NCSA's previous hybrid system, called Lincoln, and should be up and running by July 1.
Forge will combine 18 Dell PowerEdge C6145s that contain 36 nodes of dual-socket/eight-core AMD processors, with M2070 Nvidia Fermi GPU units, as well as eight Fermi units for each node, for a total of 288. Each Nvidia M2070 provides more than 500 gigaflops of performance and 6 gigabytes of memory. The input/output bandwidth of the system is expected to exceed 16 gigabytes a second.
Projects that previously used Lincoln are expected to benefit from the performance boost provided by Forge. For example, the University of Utah's Thomas Cheatham is using the supercomputer to develop better understand how proteins behave in solutions and how drugs interact with them, while NCSA's Steven Gottlieb plans to extend his work on the MILC quantum chromodynamics code to systems using multiple GPUs.
From NCSA News
View Full Article
Abstracts Copyright © 2011 Information Inc., Bethesda, Maryland, USA
No entries found