SILICON GRAPHICS INTRODUCES POWER CHALLENGEarray

Company Continues to Reinvent Supercomputing with New Distributed
Parallel Processing System

MOUNTAIN VIEW, Calif.  (Nov. 15, 1994) --  Silicon Graphics, Inc.
(NYSE:  SGI)   today  introduced  the  POWER   CHALLENGEarray,  a
distributed parallel processing system powered  by up to 144 MIPS
R8000 microprocessors to solve Grand Challenge-class problems. By
utilizing  a unique  modular approach,  the POWER  CHALLENGEarray
creates a highly-scalable supercomputing system that combines the
power of  up to  eight POWER CHALLENGE  symmetric multiprocessing
(SMP) supercomputing systems and provides  over 43 GFLOPS of peak
performance.

The POWER CHALLENGEarray combines the efficiency, flexibility and
programmability  benefits  of  Silicon  Graphics'  shared  memory
approach with up to 128 GB of  main memory, over 4 GB/sec of disk
transfer capacity and  over 63 terabytes of  standard disk space.
The POWER CHALLENGE supercomputing  systems utilized in the POWER
CHALLENGEarray shatter supercomputer price/performance levels and
scale  from two  to 18  processors,  while delivering  up to  5.4
GFLOPS of  peak performance and the  power of up to  18 Cray Y-MP
class processors in a single RISC-based system. These systems are
connected by high-performance  interconnection technology such as
HiPPI or FDDI. While the  cost of a traditional supercomputer has
restricted its use to major research laboratories and specialized
facilities, POWER  CHALLENGE systems are  designed to be  easy to
deploy   and  use,   enabling  more   scientists  and   technical
professionals to focus on solving Grand Challenge-class problems.

The  POWER CHALLENGEarray  approach is  unique among  distributed
computing  models as  each individual  system that  comprises the
array  is   itself  a  parallel  supercomputer.   This  leads  to
efficiencies  in the  computation-to-communication ratio  because
messages  are sent  between  systems for  much  larger blocks  of
parallel  processing.  The  POWER  CHALLENGEarray  is  ideal  for
solving the world's most complex supercomputing problems in areas
such  as  computational   fluid  dynamics,  structural  dynamics,
molecular dynamics,  operations research  optimization techniques
and  seismic  data  processing   projects  that  were  previously
unsolved or solved on expensive special-purpose hardware.

"With  the  introduction  of the  POWER  CHALLENGEarray,  Silicon
Graphics is continuing to redefine the rules of supercomputing by
providing deployability and accessibility to powerful technology.
The  POWER  CHALLENGEarray  maximizes  this  strategy  by  taking
advantage   of    Silicon   Graphics'    powerful   shared-memory
multiprocessing  technology  to  provide  a  very  intuitive  and
efficient method  for users to exploit  parallelism," said Forest
Baskett, Silicon  Graphics' chief  technology officer  and senior
vice president of research  and development. "The scalability and
cost-effectiveness  of the  POWER CHALLENGEarray  make it  a very
unique tool for technical professionals who are in the pursuit of
solving the Grand Challenges."

A  POWER CHALLENGEarray  can be  configured with  both the  POWER
CHALLENGE   XL  rack   system   with  two   to   18  R8000   MIPS
microprocessors  and  the  POWER  CHALLENGE  L  deskside  system,
configured  with  one to  six  R8000  microprocessors. All  POWER
CHALLENGE supercomputing  systems run Silicon Graphics'  IRIX 6.0
enhanced  64-bit  version  of  the UNIX  operating  system  which
includes a multi-threaded kernel.

The POWER CHALLENGEarray will be available by the end of Q2 1995.

-end-

Silicon  Graphics,  IRIS  and   the  Silicon  Graphics  logo  are
registered trademarks, and POWER CHALLENGE, POWER CHALLENGEarray,
Challenge,   CHALLENGEarray  IRIX   are  trademarks   of  Silicon
Graphics, Inc.
MIPS is a registered trademark of MIPS Technologies, Inc.
UNIX  is a  registered  trademark of  UNIX Systems  Laboratories,
Inc.
Cray Y-MP is a trademark of Cray Research, Inc.