currentsupercomputers

currentsupercomputers - CS6143 COMPUTER ARCHITECTURE II A...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Polytechnic Institute of NYU Page 1 of 4 Handout No : 16 December 8, 2010 A LOOK AT CURRENT SUPERCOMPUTERS CS6143 COMPUTER ARCHITECTURE II FALL 2010 Top 10 Supercomputers in the World in November 2010 From the TOP500 list (of supercomputers in the world) ( http://www.top500.org ) + Numbers indicate the performance when Linpack is run on the computers IBM BlueGene/L The original IBM BlueGene/L is a distributed-memory MIMD computer. Its hardware consists of nodes and interconnection networks. There are five types of nodes : compute , I/O, link, front- end and service . There are five interconnection networks : 3-D 64x32x32 mesh torus , global collective (combining) tree , global barrier and interrupt tree , Gigabit Ethernet and JTAG control . No Site Computer, year Cores Rmax + (TFLOPS) Rpeak (TFLOPS) Power (KWatts) 1 National Supercomputer Center, China Tianhe-1, NUDT , 2010 186368 2566.00 4001.00 4040.00 2 Oak Ridge National Laboratory Jaguar, Cray , 2009 224162 1759.00 2331.00 6950.60 3 National Supercomputer Center, China Nebulae, Dawning , 2010 120640 1271.00 2984.30 2580.00 4 GSIC Center, Japan TSUBAME, NEC/HP , 2010 73278 1192.00 2287.63 1398.61 5 DOE/SC/LBNL/NERSC Hopper, Cray , 2010 153408 1054.00 1288.63 2910.00 6 CEA, France Tera-100, Bull , 2010 138368 1050.00 1254.55 4590.00 7 DOE/NNSA/LANL Roadrunner, IBM , 2009 122400 1042.00 1375.78 2345.50 8 NICS/University of Tennessee Kraken, Cray , 2009 98928 831.70 1028.85 3090.00 9 Forschungszentrum Juelich, Germany JUGENE, IBM , 2009 294912 825.50 1002.70 2268.00 10 DOE/NNSA/LANL/SNL Cielo, Cray , 2010 107152 816.60 1028.66 2950 IBM computation chip with two PPCs A portion of the IBM BlueGene/L http://www.ibm.com/systems/deepcomputing The IBM BlueGene/L at LLNL
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Polytechnic Institute of NYU Page 2 of 4 CS6143 Handout No : 16 December 8, 2010 The largest configuration with 131072 cores at LLNL is at 367 TFLOPS and has 32TByte memory. The BlueGene runs scientific applications (physics, chemistry, biology, math, engineering,. ..). The operating system is a custom kernel on compute nodes and Linux on supporting systems. Its cost is $100M. It consumes 1.5MWatts and occupies 2500 square feet. Compute nodes run the applications. There are 65536 compute nodes each of which is a tiny computer, including a computation chip with two Power PC (PPC) CPU cores. The PPC 440 at 700MHz, has a 7-stage pipeline. There are four floating-point units (FPUs) on the node, performing 8 FP operations per clock cycle hence, the peak node speed is 5.6 GFLOPS. Each compute node has 4MByte memory. A custom lightweight kernel runs on compute nodes which
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 4

currentsupercomputers - CS6143 COMPUTER ARCHITECTURE II A...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online