High Performance Computing
Thomas Jefferson National Laboratory
1998 – 2010
Result of NP research:
NP computational studies in LQCD
Application currently being supported by:
Impact/benefit to spin-off field:
Low-cost, high performance (teraflop-class) computers based on commercial processors.
The world's first supercomputer to sustain over 1 gigaflop (one billion floating-point operations per second) was introduced by Cray Research in 1988. Now nuclear physicists are pushing the envelope, building powerful machines with supercomputer-style processor speeds without the supercomputer-style price tag. Using software designed at Jefferson Lab, in collaboration with other institutions, computer scientists and physicists are exploiting graphical processing units (GPUs) hosted in conventional computing clusters, which consist of hundreds of specially configured servers equipped with one or more high performance GPUs. The newest clusters, added in 2009 and 2010, more than quadrupled the Jefferson Lab's high-end computing capabilities and achieved more than 100 Teraflops of processing power.
The individual computing nodes in the clusters are linked together by high-speed network connections, allowing the separate processors and GPUs to essentially function as one big, compartmentalized brain. The newest clusters are continuing the work of its Jefferson Lab predecessors, running powerful computer simulations to shed light on how one of the basic forces of nature builds protons, neutrons and other particles from the basic building blocks of matter, quarks and gluons.