Looking to retain its supercomputing bragging rights, Intel Monday launched its Advanced Computing Program (ACP), a $36
million investment in future high-performance computing (HPC) designs and
software.
The Santa Clara, Calif.-based chip making giant said it is using the R&D
money to tack on more engineering resources designed to bolster next
generation mainframes and clusters, the kind you usually see on the biannual
supercomputer TOP500 list.
Since the list was updated Monday, Intel scored the most of any other chipmaker including RISC-based
“We are ecstatic about the acceptance of Intel’s standards-based
approach,” Intel senior vice president Mike Fister said in a statement. “Our focus on this market segment and its future will benefit from our investment and work with key academia, government and industry advocates on the Advanced Computing Program effort.”
That three-year program will bring scientists from Intel together with
leaders in the high-performance computing community to develop and prototype
advanced computing technologies. The No. 1 chipmaker said these technologies
include new architectures, advanced system software, and software tools,
languages, and libraries. Ultimately the ACP technologies will be deployed
to the industry via next-generation “off-the-shelf” computing building
blocks, such as the Intel Xeon, Intel Itanium 2, and Intel Pentium 4
processors, along with ACP system recommendations.
“The Advanced Computing Program is a response to the need for more
balanced high performance systems that have greater capabilities and better
programming environments for end users,” said Justin Rattner, Intel senior
fellow and director of the Intel Systems Technology Lab. “Intel is excited
to help deliver a qualitative change in the way scientists and researchers
will soon be able to do their work. We also intend to work with the industry
to make sure that innovations developed at the high-end of computing will
eventually benefit all computer users.”
Other prominent Intel based systems on the TOP500 list include the
fourth-ranked system at the National Center for Supercomputing Applications
(NCSA) using 2,500 Intel Xeon processors, and the fifth-ranked system at
Pacific Northwest National Laboratory using 1,936 Intel Itanium 2
processors. In addition, the seventh- and 10th-ranked systems on the list
are at the University of California at Lawrence Livermore National
Laboratory (LLNL) and use 2,304 and 1,920 Intel Xeon processors
respectively. The 14th-ranked system at the Chinese Academy of Science uses
1,024 Intel Itanium 2 processors.
In a related announcement yesterday, Intel and LLNL disclosed that they
will deploy a supercomputer in January that will use nearly 4,000 Intel
Itanium 2 processors and is expected to be a top-ranked system included in
the next list to be issued in June.
Also as part of its HPC push, Intel demonstrated a cluster setup that
delivers over one teraflop of computing power using industry standard
building blocks including InfiniBand and PCI Express. The 192-node
configuration, presented at the Supercomputing 2003 show in Phoenix, was
assembled in less than two days on the show floor. Historically, teraflop
configurations have taken months to configure and have not been possible to
demonstrate in a show environment.
