Fermilab’s Amitoj Singh and Don Holmgren examine one of the new GPUs used for lattice QCD calculations. Photo: Brad Hooker
The latest addition to computing power at U.S. Department of
Energy’s Fermi National Accelerator Laboratory is a 45-teraflop cluster of
graphics processing units (GPU) that scientists use to explore the properties
of the strong nuclear force. The GPU nodes power through data faster than any
other computing nodes at more than five times the rate of the processing units
of the previous generation.
The cluster is part of a national project called USQCD.
Quantum chromodynamics, or QCD, is the theory that explains the properties and
behavior of quarks and gluons. Scientists compute the particles’ subatomic
interaction, the strong nuclear force, using algorithms and techniques known as
lattice QCD. The USQCD collaboration develops the software and hardware needed
to meet the high demands of lattice QCD calculations, which require tens of
thousands of processors.
While industry would like to see increasingly powerful
processors applied to cell phones, laptops and other consumer electronics, the
USQCD collaboration aims to determine how important GPUs will be to scientific
“We don’t know if GPU-like chips are the way of the
future or just a flash in the pan now,” said Fermilab physicist Paul
Mackenzie, spokesman for the national collaboration of QCD, in a recent
interview. “The scientific computing world is changing. Computers 10 years
from now will look very different from how they’ve looked the last 10 years.”