Go Back
Click on home button to open mainpage

Earth Simulator System

2002, NEC Japan


birdview of ESS installation



Dr. Mitsuo Yokokawa chief designer

 

papers & manuals
see above

hardware
see above

software
see above

keywords
supercomputers, vector processors, nec

Introduction

The Earth Simulator is a special purpose machine, made by NEC with the same type of vector technology as is available on the SX-6. The decision of NEC to base the design entirely on vector processors indicates a new trend towards supercomputer design. Vector processors are already around for some time but no manufacturer took the risk of implementing them full scale in supercomputers. Until the ESS came along.

This is a different approach than the Americans are taking, who base the large machines on off-the shelf technology. A clear advantage of the Japanese approach is that they "only" need 5120 processors for the Earth Simulator. The Asci White at the US Lawrence Livermore National Laboratory and made by IBM, already needs 8192 processors for 7 Tflop/s.


screendump of ESS

The importance of this machine can be found in the type of research: environment. The issues of environmental changes has become more and more important during the past decade where pollution became a serious problem threatening the climate and our world we live in. And the climate is exactly where the ESS is going to support research in.

 

Description(1)

 


building layout (Apr. 4 tennis courts)(1)

The ultra high-speed parallel computing system, "Earth Simulator" attained the best computing performance in the world according to the Linpack benchmark test. The result of this test of execution performance, 35.61 TFlops (trillion operations per second), was approved and the Earth Simulator was registered as the world' s fastest supercomputer by Dr. Jack J. Dongarra of the University of Tennessee, USA, who announced the world Linpack benchmark test results.


Part of machine installation ESS(1)

The Earth Simulator consists of 640 supercomputers that are connected by a high-speed network (data transfer speed; 12.3 GBytes). Each supercomputer (1 node) contains eight vector processors with a peak performance of 8GFlops and a high-speed memory of 16 GBytes. The total number of processors is 5120 (8 x 640), which translates to a total of approximately 40 TFlops peak performance, and a total main memory of 10 TeraBytes.


NEC's ESS vector processor(1)

35.61 TFlops is achieved from the operation of 638 nodes (5,104 processors) and an efficiency performance percentage of 87.2%.

According to the TOP500 list of supercomputers in the world, the second is the ASCI White system in the US, whose performance is 7.266 TFlops (peak performance 12.288 TFlops, peak performance percentage 58.8%). The Earth Simulator has five times the capacity of the US system.



Top 500 performance chart(1)


Specifications (6)

Earth-Simulator runs at 35860.00 GFlops and is installed at Earth Simulator Center Japan Research it contains 5120 vector processors and peaks at 40960.00 GFlops

Based on the NEC SX architecture, 640 nodes, each node with 8 vector processors (8 Gflop/s peak per processor), 2 ns cycle time, 16GB shared memory. Total of 5120 total processors, 40 TFlop/s peak, and 10 TB memory.
It has a single stage crossbar (1800 miles of cable) 83,000 copper cables, 16 GB/s cross section bandwidth.
700 TB disk space
1.6 PB mass store
Area of computer = 4 tennis courts, 3 floors
Rmax from LINPACK MPP Benchmark Ax=b, dense problem
Linpack Benchmark = 35.6 TFlop/s
Problem of size n = 1,041,216; (8.7 TB of memory)
Half of peak (n ½ ) achieved at n ½ = 265,408
Benchmark took 5.8 hours to run.
Software: for the most part Fortran using MPI

Comparison with other supers:

  • Sum of all the DOE computers = 24 TFlop/s
  • Performance of ESC > (Top 18 Computers)
  • Performance of ESC > (Top 20 Computers in the US)
  • Performance of ESC > All the DOE and DOD machines (27.6 TFlop/s)
  • Performance of ESC >> the 3 NSF Center's computers (8.4 TFlop/s)


Release history

1997 design started
1998 research and development to harness hardware finished
1999 manufacturing started end 1999
2001 manufacturing ready
2001 late 2001 in service
2002 first announcement

 

bar

Go Back Last Updated on 23 September, 2002 For suggestions please mail the editors 


Footnotes & References

1 all pictures and main text republished with permission of NEC inc.
2 1 TFLOPS is a term for a trillion floating-point operations per second."Tera" is a prefix for one trillion (ten raised to the twelfth power),"giga" for one billion (ten raised to the tenth power). When the Linpack program was executed, the Earth Simulator achieved 35.61 TFLOPS, which is 35.61 trillion floating-point operations per second on average.
3 Linpack benchmark test:
A numerically intensive test that has been widely used to measure and compare the processing performance of supercomputers, and similar devices. Propounded by Dr. J. Dongarra of the University of Tennessee, the test measures the processing performance (in Gigaflops) of a computer program that solves a large-scale linear equation.

4 TOP500: A web site listing the sites that have the 500 most powerful computer systems installed. The best Linpack benchmark performance achieved is used as a performance measure in ranking the computers. The TOP500 list has been updated twice every year since 1993. http://www.top500.org/
5 Vector Processor: A processor that executes computational operations at high speed with a single instruction. Vector processors provide high-sustained performance.
6 techincal data quoted from: www.top500.org