USC 76-Node Macintosh Cluster Achieves 233 GigaFlops
Large Macintosh cluster at USC achieves over 1/5 TeraFlop on 152 G4's and
demonstrates excellent scalability.
Los Angeles, California, USA - January 11, 2002 -
Researchers and staff at the University of Southern California (USC)
collaborated with scientists from the University of California, Los Angeles, (UCLA)
to achieve a milestone for cluster computing and the Macintosh platform.
Software from Dauger Research, Inc., and UCLA achieved 233 Gigaflops on a
Macintosh cluster at USC's Language Arts computer lab and other USC facilities.
The software used to perform this calculation is the latest incarnation of
project begun by a pair of UCLA physics researchers in 1998. Their
software is used world-wide for easy-to-use and numerically-intensive
parallel computing. This is the first time, however, that it has been
run on 76 nodes or has achieved over 1/5 TeraFlop (1 TF = 1000 GF =
one trillion floating-point operations per second).
In addition, no evidence could be found for an intrinsic limit to the size of
a Macintosh cluster, which indicates that Mac clusters have excellent potential
The hardware consisted of 56 Power Mac Dual-Processor G4/533's plus 20
Power Mac Dual-Processor G4/450's connected via a large 100BaseT switch.
The application run was
the AltiVec Fractal Carbon demo, which used
for its message passing. The parallel computing job was launched using
and the operating system was
Mac OS 9.2.1.
This benchmark was performed in December 2001, while USC was on Christmas break.
Now that USC is back in session, the nodes have returned to work as a computer lab for
their Language Arts department.
Full details are available on
Dauger Research web site.
Dauger Research, Inc., was incorporated in September 2001 and founded by
Dr. Dean E. Dauger. Dr. Dauger is the award-winning author of Atom in a
Box and Fresnel Diffraction Explorer and has co-authored award-winning
commercial software as early as 1992. After completing his Ph. D. in
physics, he founded Dauger Research, Inc., to bridge the divides between
the scientifically and technically complex and the mainstream by making
high-performance computation and visualization easy to use and accessible