Overview Overview Watson Research Center Self Adapting Numerical - PDF document

Overview Overview Watson Research Center Self Adapting Numerical Self Adapting Numerical Software (SANS) – – Effort and Effort and Software (SANS) ♦ Quick look at fastest computers Fault Tolerance in Linear Fault Tolerance in Linear � From the November Top500 ♦ Techniques for fault tolerant Algebra Algorithms Algebra Algorithms computations for iterative methods Jack Dongarra � Strategies when we start to using 10’s University of Tennessee of thousands of processors and Oak Ridge National Laboratory 00 2 1/25/2005 1 TOP500 Performance – – November 2004 November 2004 TOP500 Performance H. Meuer, H. Simon, E. Strohmaier, & JD H. Meuer, H. Simon, E. Strohmaier, & JD 1.127 PF/s - Listing of the 500 most powerful 1 Pflop/ s IBM Computers in the World BlueGene/L 100 Tflop/ s SUM 70.72 TF/s - Yardstick: Rmax from LINPACK MPP NEC 10 Tflop/ s 1.167 TF/s N=1 Earth Simulator IBM ASCI White Ax=b, dense problem 1 Tflop/ s LLNL 850 GF/s TPP performance 59.7 GF/s Intel ASCI Red Sandia 100 Gflop/ s Rate Fuj itsu 10 Gflop/ s 'NWT' NAL N=500 My Laptop - Updated twice a year 0.4 GF/s Size 1 Gflop/ s SC‘xy in the States in November 100 Mflop/ s 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Meeting in Mannheim, Germany in June - All data available from www.top500.org 00 3 00 4 24th List: The TOP10 Architectures / Systems 24th List: The TOP10 Architectures / Systems Rmax Manufacturer Computer Installation Site Country Year #Proc [TF/s] 500 BlueGene/L 1 IBM 70.72 DOE/IBM USA 2004 32768 β -System S IMD Columbia 2 SGI 51.87 NASA Ames USA 2004 10160 400 Altix, Infiniband S ingle Proc. 3 NEC Earth-Simulator 35.86 Earth Simulator Center Japan 2002 5120 300 MareNostrum Barcelona Supercomputer Cluster 4 IBM 20.53 Spain 2004 3564 BladeCenter JS20, Center Myrinet Thunder Lawrence Livermore 200 Constellation 5 CCD 19.94 USA 2004 4096 National Laboratory Itanium2, Quadrics Los Alamos ASCI Q 6 HP 13.88 USA 2002 8192 National Laboratory AlphaServer SC, Quadrics S MP 100 X 7 Self Made 12.25 Virginia Tech USA 2004 2200 Apple XServe, Infiniband MPP BlueGene/L Lawrence Livermore 0 8 IBM/LLNL 11.68 USA 2004 8192 DD1 500 MHz National Laboratory 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Naval Oceanographic 9 IBM pSeries 655 10.31 USA 2004 2944 Office Tungsten 10 Dell 9.82 NCSA USA 2003 2500 PowerEdge, Myrinet 00 5 00 6 399 system > 1 TFlop/s; 294 machines are clusters, top10 average 8K proc #39 NCAR 1

Top500 Performance by Manufacture (11/04) Processor Types Processor Types Sun Cray Hitachi 0% 2% 1% Fujitsu 2% Intel NEC 0% 4% 500 SGI 7% S IMD 400 Vector others S calar 14% IBM 300 49% S parc MIPS 200 intel HP HP 100 21% Power 0 Alpha 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 00 7 00 8 Interconnects / Systems Interconnects / Systems Fuel Efficiency: Gflops Fuel Efficiency: Gflops/Watt /Watt 0.9 0.8 500 0.7 Gflops/Watt 0.6 Others 400 0.5 Infiniband 0.4 Quadrics 300 0.3 Gigabit Ethernet 0.2 Cray Int erconnect 200 0.1 Myrinet 0 S P S witch 1100 Dual 2.3 GHz Apple XServe/Mellanox Infiniband 4X/Cisco GigE SGI Altix 1.5 GHz, Voltaire Infiniband ASCI Q - AlphaServer SC45, 1.25 GHz ASCI White, SP Power3 375 MHz Cray X1 BlueGene/L DD2 beta-System (0.7 GHz PowerPC 440) BlueGene/L DD1 Prototype (0.5GHz PowerPC 440 w/Custom) Intel Itanium2 Tiger4 1.4GHz - Quadrics Earth-Simulator PowerEdge 1750, P4 Xeon 3.06 GHz, Myrinet BlueGene/L DD2 Prototype (0.7 GHz PowerPC 440) LNX Cluster, Xeon 3.4 GHz, Myrinet RIKEN Super Combined Cluster Integrity rx2600 Itanium2 1.5 GHz, Quadrics Dawning 4000A, Opteron 2.2 GHz, Myrinet MCR Linux Cluster Xeon 2.4 GHz - Quadrics Opteron 2 GHz, Myrinet TeraGrid, Itanium2 1.3/1.5 GHZ, Myrinet SP Power3 375 MHz 16 way xSeries Cluster Xeon 2.4 GHz - Quadrics eServer Opteron 2.2 GHz. Myrinet AIST Super Cluster P-32, Opteron 2.0 GHz, Myrinet xSeries Xeon 3.06 GHz, Quadrics 100 eServer BladeCenter JS20+ (PowerPC970 2.2 GHz), Myrinet eServer pSeries 655 (1.7 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) eServer pSeries 655/690 (1.5/1.7 Ghz Power4+) eServer pSeries 690 (1.7 GHz Power4+) eServer pSeries 690 (1.7 GHz Power4+) Crossbar N/ A 0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Top 20 systems 00 9 00 10 Based on processor power rating only Fuel Efficiency: Gflops Gflops/Watt /Watt Fuel Efficiency: Gflops Gflops/Watt /Watt Fuel Efficiency: Fuel Efficiency: 0.9 0.9 0.8 0.8 0.7 0.7 Gflops/Watt Gflops/Watt 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 SGI Altix 1.5 GHz, Voltaire Infiniband ASCI Q - AlphaServer SC45, 1.25 GHz Cray X1 SGI Altix 1.5 GHz, Voltaire Infiniband ASCI Q - AlphaServer SC45, 1.25 GHz Cray X1 BlueGene/L DD2 beta-System (0.7 GHz PowerPC 440) 1100 Dual 2.3 GHz Apple XServe/Mellanox Infiniband 4X/Cisco GigE Intel Itanium2 Tiger4 1.4GHz - Quadrics Earth-Simulator PowerEdge 1750, P4 Xeon 3.06 GHz, Myrinet BlueGene/L DD2 Prototype (0.7 GHz PowerPC 440) LNX Cluster, Xeon 3.4 GHz, Myrinet RIKEN Super Combined Cluster Integrity rx2600 Itanium2 1.5 GHz, Quadrics Dawning 4000A, Opteron 2.2 GHz, Myrinet MCR Linux Cluster Xeon 2.4 GHz - Quadrics Opteron 2 GHz, Myrinet ASCI White, SP Power3 375 MHz TeraGrid, Itanium2 1.3/1.5 GHZ, Myrinet SP Power3 375 MHz 16 way xSeries Xeon 3.06 GHz, Quadrics BlueGene/L DD2 beta-System (0.7 GHz PowerPC 440) 1100 Dual 2.3 GHz Apple XServe/Mellanox Infiniband 4X/Cisco GigE Intel Itanium2 Tiger4 1.4GHz - Quadrics Earth-Simulator PowerEdge 1750, P4 Xeon 3.06 GHz, Myrinet BlueGene/L DD2 Prototype (0.7 GHz PowerPC 440) LNX Cluster, Xeon 3.4 GHz, Myrinet RIKEN Super Combined Cluster Integrity rx2600 Itanium2 1.5 GHz, Quadrics Dawning 4000A, Opteron 2.2 GHz, Myrinet MCR Linux Cluster Xeon 2.4 GHz - Quadrics Opteron 2 GHz, Myrinet ASCI White, SP Power3 375 MHz TeraGrid, Itanium2 1.3/1.5 GHZ, Myrinet SP Power3 375 MHz 16 way xSeries Xeon 3.06 GHz, Quadrics eServer BladeCenter JS20+ (PowerPC970 2.2 GHz), Myrinet BlueGene/L DD1 Prototype (0.5GHz PowerPC 440 w/Custom) eServer pSeries 655 (1.7 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) xSeries Cluster Xeon 2.4 GHz - Quadrics eServer pSeries 655/690 (1.5/1.7 Ghz Power4+) eServer Opteron 2.2 GHz. Myrinet AIST Super Cluster P-32, Opteron 2.0 GHz, Myrinet eServer pSeries 690 (1.7 GHz Power4+) eServer pSeries 690 (1.7 GHz Power4+) eServer BladeCenter JS20+ (PowerPC970 2.2 GHz), Myrinet BlueGene/L DD1 Prototype (0.5GHz PowerPC 440 w/Custom) eServer pSeries 655 (1.7 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) eServer pSeries 690 (1.9 GHz Power4+) xSeries Cluster Xeon 2.4 GHz - Quadrics eServer pSeries 655/690 (1.5/1.7 Ghz Power4+) eServer Opteron 2.2 GHz. Myrinet AIST Super Cluster P-32, Opteron 2.0 GHz, Myrinet eServer pSeries 690 (1.7 GHz Power4+) eServer pSeries 690 (1.7 GHz Power4+) Top 20 systems Top 20 systems 00 11 00 12 Based on processor power rating only Based on processor power rating only 2