NCSA installing 153 teraflop supercomputer
05.18.11 - Permalink
The National Center for Supercomputing Applications has begun installation of a 153 teraflop supercomputer, dubbed Forge. Forge is a hybrid system that combines both CPUs and general purpose graphics-processing units (GPUs); it will replace NCSA's previous hybrid system, known as Lincoln, and is expected to enter production 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 housed in Dell's C410x PCI expansion enclosures; there are eight Fermi units for each node, for a total of 288. Each NVIDIA M2070 provides more than 500 gigaflops of double-precision performance and 6GB of GDDR5 memory.
A QDR InfiniBand interconnect fabric will interconnect the nodes; 700 terabytes of GPFS filesystem space will be provided by two Data Direct Networks SFA 10000 units. The I/O bandwidth of the system is expected to exceed 16GB/sec.
Projects that have used Lincoln in the past will have the opportunity to continue with Forge. For example,
- Klaus Schulten's research team, which uses NAMD to study the organization and function of proteins and protein complexes within cells, found that two of Lincoln's GPUs were equivalent of 24 of Lincoln's CPU cores, while eight of its GPUs were equivalent to 96 CPU cores.
- 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.
- During a sabbatical at NCSA, Steven Gottlieblone of the principal architects of the MILC quantum chromodynamics codedeveloped a single GPU version of several important components of the code, and this work is now being extended to mullti-GPUs. (Read more about porting MILC to GPUs)
Forge is being installed at the University of Illinois' new National Petascale Computing Facility, which offers a state-of-the-art 20,000-square-foot machine room. The system will be allocated through the National Science Foundation TRAC process.