Skip to main content

IBM Power Systems performance benchmarks

Including performance benchmarks for Power servers, System p, and System i

Power is workload optimizing system

IBM Power™ Systems demonstrate IBM's continued commitment to organizations of all sizes. In addition to superior economics and breakthrough energy efficiency and virtualization technologies, Power systems offer outstanding performance. In fact, IBM Power Systems took the lead in over 100 key computing performance benchmarks. But don't take our word for it. Read the results below.

Tab navigation

LINPACK

Measures how fast a dedicated system can solve a dense system of linear equations as a means of determining the floating-point performance of the system. It is used to gauge effectiveness for high performance and technical computing.

  • A 1-core IBM Power 570 (5 GHz) is the best 1-core HPC result (17.47 Gigaflop/sec).
  • A 4-core IBM Power 520 (4.7 GHz) is the best 4-core HPC result (65.01 Gigaflop/sec).
  • A 4-core IBM Power 520 (4.7 GHz) is the best 4-core Linux RISC SMP HPC result (64.42 Gigaflop/sec).
  • An 8-core IBM Power 550 (5 GHz) is the best 8-core HPC result (137.6 Gigaflop/sec).
  • An 8-core IBM Power 550 (5 GHz) is the best 8-core Linux RISC SMP HPC result (137.1 Gigaflop/sec).
  • A 12-core IBM Power 730 (3.72 GHz) is the best 12-core HPC result (335.2 Gigaflop/sec).
  • A 16-core IBM Power 780 (3.86 GHz) is the best 16-core HPC result (445 Gigaflop/sec).
  • A 16-core IBM Power 740 (3.55 GHz) is the best 16-core Linux HPC result (435.8 Gigaflop/sec).
  • A 32-core IBM Power 780 (4.14 GHz) is the best 32-core HPC result (944 Gigaflop/sec).
  • A 32-core IBM Power 750 Express (3.55 GHz) is the best 32-core Linux HPC result (870 Gigaflop/sec).
  • A 64-core IBM Power 780 (3.86 GHz) is the best 64-core HPC result (1,775 Gigaflop/sec).
  • A 64-core IBM Power 780 (3.86 GHz) is the best 64-core Linux HPC result (1,772 Gigaflop/sec).
  • A 96-core IBM Power 780 (3.44 GHz) is the best 96-core HPC result (2512 Gigaflop/sec, Linux).
  • A 128-core IBM Power 795 (4.25 GHz) is the best 128-core HPC result (3,784 Gigaflop/sec, Linux).
  • A 256-core IBM Power 795 (4.00 GHz) is the best 256-core HPC result (6,902 Gigaflop/sec, Linux).

Source: http://performance.netlib.org/performance/html/PDSreports.html (link resides outside of ibm.com)

Stream

Measures sustainable memory bandwidth and the corresponding computation rate for simple vector kernels.

  • A 4-core IBM System p5 570 (1.9 GHz) is the best 4-core result (26,214 MB/s Triad tuned).
  • An 8-core IBM System p5 575 (2.2 GHz) is the best 8-core result (100,523 MB/s Triad tuned).
  • A 16-core IBM System p5 575 (1.9 GHz) is the best 16-core result (86,379 MB/s Triad tuned).
  • A 24-core IBM Power 750 Express (3.30 GHz) is the best 24-core result (122,766 MB/s Triad standard, Linux).
  • A 64-core IBM Power 595 (5 GHz) is the best 64-core result (805,804 MB/s Triad standard).

Source: http://www.cs.virginia.edu/stream/ (link resides outside of ibm.com)

SPECOMPM2001

Measures a system's parallel processing capabilities for medium problem sizes using a suite of applications based on the OpenMP standards for shared-memory parallel processing. It is used to gauge effectiveness for high performance and technical computing.

  • A 2-core IBM BladeCenter JS12 Express (3.8 GHz) running Linux is the best 2-core (1 chip, 4 threads) result (12,885 SPECompMpeak2001).
  • A 4-core IBM Power 520 (4.2 GHz) running Linux is the best 4-core (2 chip, 8 threads) result (20,443 SPECompMpeak2001).
  • A 12-core IBM Power 730 (3.72 GHz) running Linux is the best 12-core (2 chip, 48 threads) result (80,952 SPECompMpeak2001).
  • A 16-core IBM Power 740 (3.55 GHz) running Linux is the best 16-core (2 chip, 64 threads) result (95,002 SPECompMpeak2001
  • A 64-core IBM Power 595 (5 GHz) running Linux is the best overall (32 chip, 128 threads) result (242,116 SPECompMpeak2001).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECOMPL2001

Measures a system's parallel processing capabilities for large problem sizes using a suite of applications based on the OpenMP standards for shared-memory parallel processing. It is used to gauge effectiveness for high performance and technical computing.

  • A 32-core IBM Power 750 Express (3.55 GHz) is the best 32-core (4 chips, 128 threads) result (773,674 SPECompLpeak2001, Linux).
  • A 64-core IBM Power 780 (3.86 GHz) is the best 64-core (8 chips, 256 threads) result (1,797,211 SPECompLpeak2001, Linux).

Source: http://www.spec.org (link resides outside of ibm.com)

Back to top

SPECfp2006

Measures the compute-intensive floating point speed of the computer processor, memory architecture, and compilers.

  • A 1-core IBM Power 595 (5 GHz) is the best 1-core result (24.9 SPECfp2006 result, 1 core, 1 chip, 2 cores/chip, 1 thread/core).
  • A 1-core IBM Power 570 (4.7 GHz) is the best 1-core Linux result (22.5 SPECfp2006 result, 1 core, 1 chip, 2 cores/chip, 1 thread/core, Linux).
  • A 16-core IBM Power 780 (3.86 GHz) is the best server result in the industry (71.5 SPECfp2006 result, 16 cores, 2 chips, 8 cores/chip, 1 thread/core).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECfp_rate2006

Determines the number of tasks from a collection of 17 floating-point compute intensive programs that can be run in a specified time.

  • A 6-core IBM Power 710 (3.72 GHz) is the best 6-core system (213 SPECfp_rate2006 result, 6 cores, 1 chip, 6 cores/chip, 4 threads/core).
  • An 8-core IBM Power 710 (3.55 GHz) is the best 8-core system (248 SPECfp_rate2006 result, 8 cores, 1 chip, 8 cores/chip, 4 threads/core).
  • A 12-core IBM Power 730 (3.72 GHz) is the best 12-core system (423 SPECfp_rate2006 result, 12 cores, 2 chips, 6 cores/chip, 4 threads/core).
  • A 16-core IBM Power 740 (3.55 GHz) is the best 16-core Linux system (497 SPECfp_rate2006 result, 16 cores, 2 chips, 8 cores/chip, 4 threads/core).
  • A 16-core IBM Power 780 (3.86 GHz) is the best 16-core system (586 SPECfp_rate2006 result, 16 cores, 2 chips, 8 cores/chip, 4 threads/core).
  • A 32-core IBM Power 780 (4.14 GHz) is the best 32-core result (1310 SPECfp_rate2006 result, 32 cores, 8 chips, 4 cores/chip, 4 threads/core, Linux).
  • A 48-core IBM Power 770 (3.50 GHz) is the best 48-core result (1,760 SPECfp_rate2006 result, 48 cores, 8 chips, 6 cores/chip, 4 threads/core).
  • A 64-core IBM Power 780 (3.86 GHz) is the best 64-core Linux system (2,550 SPECfp_rate2006 result, 64 cores, 8 chips, 8 cores/chip, 4 threads/core).
  • A 64-core IBM Power 780 (3.92 GHz) is the best 64-core system (2,640 SPECfp_rate2006 result, 64 cores, 8 chips, 8 cores/chip, 4 threads/core).
  • A 96-core IBM Power 780 (3.44 GHz) is the best 96-core Linux system (3,080 SPECfp_rate2006 result, 96 cores, 16 chips, 6 cores/chip, 4 threads/core).
  • A 96-core IBM Power 780 (3.44 GHz) is the best 96-core system (3,150 SPECfp_rate2006 result, 96 cores, 16 chips, 6 cores/chip, 4 threads/core).
  • A 128-core IBM Power 795 (4.25 GHz) is the best 128-core Linux system (5,860 SPECfp_rate2006 result, 128 cores, 16 chips, 8 cores/chip, 4 threads/core).
  • A 128-core IBM Power 795 (4.25 GHz) is the best 128-core system (5,870 SPECfp_rate2006 result, 128 cores, 16 chips, 8 cores/chip, 4 threads/core).
  • A 256-core IBM Power 795 (4.00 GHz) is the best 256-core system (10,500 SPECfp_rate2006 result, 256 cores, 32 chips, 8 cores/chip, 4 threads/core).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECint2006

Measures the compute-intensive integer speed of the computer processor, memory architecture, and compilers.

  • A 1-core IBM Power 570 (4.7 GHz) ranked #1 1-core Linux result in the industry (21.7 SPECint2006 result, 1 core, 1 chip, 2 cores/chip, 1 thread/core, Linux).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECint_rate2006

Determines the number of tasks from a collection of 12 integer compute intensive programs that can be run in a specified time.

  • A 6-core IBM Power 710 (3.72 GHz) is the best 6-core system (239 SPECint_rate2006 result, 6 cores, 1 chip, 6 cores/chip, 4 threads/core).
  • A 12-core IBM Power 730 (3.72 GHz) is the best 12-core system (476 SPECint_rate2006 result, 12 cores, 2 chips, 6 cores/chip, 4 threads/core).
  • A 16-core IBM Power 740 (3.55 GHz) is the best 16-core Linux system (581 SPECint_rate2006 result, 16 cores, 2 chips, 8 cores/chip, 4 threads/core).
  • A 16-core IBM Power 780 (3.86 GHz) is the best 16-core system (652 SPECint_rate2006 result, 16 cores, 2 chips, 8 cores/chip, 4 threads/core).
  • A 32-core IBM Power 780 (4.14 GHz) is the best 32-core system (1,470 SPECint_rate2006 result, 32 cores, 8 chips, 4 cores/chip, 4 threads/core, Linux).
  • A 48-core IBM Power 770 (3.50 GHz) is the best 48-core system (1,930 SPECint_rate2006 result, 48 cores, 8 chips, 6 cores/chip, 4 threads/core).
  • A 64-core IBM Power 780 (3.86 GHz) is the best 64-core Linux system (2,740 SPECint_rate2006 result, 64 cores, 8 chips, 8 cores/chip, 4 threads/core).
  • A 64-core IBM Power 780 (3.92 GHz) is the best 64-core system (2,770 SPECint_rate2006 result, 64 cores, 8 chips, 8 cores/chip, 4 threads/core).
  • A 96-core IBM Power 780 (3.44 GHz) is the best 96-core system (3,560 SPECint_rate2006 result, 96 cores, 16 chips, 6 cores/chip, 4 threads/core, Linux).
  • A 128-core IBM Power 795 (4.25 GHz) is the best 128-core Linux system (6,130 SPECint_rate2006 result, 128 cores, 16 chips, 8 cores/chip, 4 threads/core).
  • A 128-core IBM Power 795 (4.25 GHz) is the best 128-core system (6,150 SPECint_rate2006 result, 128 cores, 16 chips, 8 cores/chip, 4 threads/core).
  • A 256-core IBM Power 795 (4.00 GHz) is the best 256-core system (11,300) SPECint_rate2006 result, 256 cores, 32 chips, 8 cores/chip, 4 threads/core, Linux).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECfp2000

Measures how fast a single processor, cache, and memory can run a collection of 14 floating-point compute-intensive programs.

  • A 1-core IBM System p5 595 (2.3 GHz) ranked #1 in the industry (3,642 SPECfp2000 result).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECfp_rate2000

Determines the number of tasks from a collection of 14 floating-point compute intensive programs that can be run in a specified time.

  • A 2-core IBM System p5 520 (1.9 GHz) is the best 2-core system (67.6 SPECfp_rate2000 result).
  • A 4-core IBM System p5 550 (2.1 GHz) is the best 4-core system (149 SPECfp_rate2000 result).
  • An 8-core IBM System p5 575 (2.2 GHz) is the best 8-core system (382 SPECfp_rate2000 result).
  • A 16-core IBM System p5 575 (1.9 GHz) is the best 16-core system (571 SPECfp_rate2000 result).
  • A 32-core IBM eServer p5 590 (1.65 GHz) is the best 32-core system (870 SPECfp_rate2000 result).
  • A 64-core IBM System p5 595 (2.3 GHz) is the best 64-core system (2,406 SPECfp_rate2000 result).

Source: http://www.spec.org (link resides outside of ibm.com)

SPECint_rate2000

Determines the number of tasks from a collection of 12 integer compute intensive programs that can be run in a specified time.

  • A 64-core IBM System p5 595 (2.3 GHz) is the best 64-core system (1,513 SPECint_rate2000 result).

Source: http://www.spec.org (link resides outside of ibm.com)

Back to top

NAMD V2.7b1

NAMD is a parallel molecular dynamics code designed for high-performance simulation of large biomolecular systems. NAMD was designed to run efficiently on such parallel machines for simulating large molecules. Based on Charm++ parallel objects, NAMD is a simulation engine implementing the molecular dynamics algorithm as applied to biomolecular system with higher optimization and efficiency. NAMD was developed by the Theoretical and Computational Biophysics Group in the Beckman Institute for Advanced Science and Technology at the University of Illinois at Urbana-Champaign.

  • A 1-core IBM Power 755 (3.30 GHz) is the best 1-core system (0.7195 seconds per apoA1 step, 2.2502 seconds per f1 ATPase step).
  • A 2-core IBM Power 755 (3.30 GHz) is the best 2-core system (0.3567 seconds per apoA1 step, 1.1254 seconds per f1 ATPase step, 3.9473 seconds per STMV step).
  • An 4-core IBM Power 755 (3.30 GHz) is the best 4-core system (0.1822 seconds per apoA1 step, 0.5733 seconds per f1 ATPase step, 1.9566 seconds per STMV step).
  • An 8-core IBM Power 755 (3.30 GHz) is the best 8-core system (0.0943 seconds per apoA1 step, 0.2906 seconds per f1 ATPase step, 0.9975 seconds per STMV step).
  • A 16-core IBM Power 755 (3.30 GHz) is the best 16-core system (0.0473 seconds per apoA1 step, 0.1485 seconds per f1 ATPase step, 0.5069 seconds per STMV step).
  • A 32-core IBM Power 755 (3.30 GHz) is the best 32-core system (0.0249 seconds per apoA1 step, 0.742 seconds per f1 ATPase step, 0.2516 seconds per STMV step).
  • A 64-core IBM Power 755 cluster (3.30 GHz) is the best 64-core system (0.0135 seconds per apoA1 step, 0.0379 seconds per f1 ATPase step, 0.1287 seconds per STMV step).
  • A 128-core IBM Power 755 cluster (3.30 GHz) is the best 128-core system (0.0070 seconds per apoA1 step, 0.0196 seconds per f1 ATPase step, 0.0665 seconds per STMV step).

Source: http://www.ks.uiuc.edu/Research/namd/ (link resides outside of ibm.com)

SIMULIA

Abaqus is SIMULIA's flagship product and provides a comprehensive, general-purpose finite element analysis tool that includes a variety of time and frequency-domain analysis procedures. A lower number indicates faster performance.

Abaqus/Standard Version 6.6

  • A 1-core IBM Power 575 (4.7 GHz) shows best 1-core Abaqus/Standard performance (elapsed times for s2a:2318, s2b:1882, s3a:446, s3b:2369, s3c:2077, s4b:8450, s4c:4592).
  • A 2-core IBM Power 575 (4.7 GHz) shows best 2-core Abaqus/Standard performance (elapsed times for s2a:1231, s2b:1008, s4b:4745, s4c:2933).
  • A 4-core IBM Power 575 (4.7 GHz) shows best 4-core Abaqus/Standard performance (elapsed times for s2a:683, s2b:564, s4b:2912, s4c:1941,s6:3185).
  • An 8-core IBM Power 575 (4.7 GHz) shows best 8-core Abaqus/Standard performance (elapsed times for s2b:353, s4a:340, s4b:2002, s4c:1509).
  • A 16-core IBM Power 575 (4.7 GHz) shows best 16-core Abaqus/Standard performance (elapsed times for s2b:262, s4b:1608, s4c: 1534).

Abaqus/Explicit Version 6.6

  • A 4-core IBM System p5 570 (2.2 GHz) best 4-core Abaqus/Explicit performance (elapsed time for e3:3033).
  • An 8-core IBM System p5 570 (2.2 GHz) shows best 8-core Abaqus/Explicit performance (elapsed time for e3:1556).

Source: http://www.simulia.com/support/sup_systems_info.html (link resides outside of ibm.com)

ANSYS — ANSYS V11

ANSYS is a Finite Element Analysis Simulation Software package that provides solutions for conceptual design through final stage testing and performance validation from design concept to final-stage testing and performance validation. The ANSYS Product Suite is used in Structural, Thermal, Mechanical, Acoustics, Computational Fluid Dynamics (CFD), Electrical and Electromagnetic Analyses

  • A 1-core IBM Power 550 (4.2 GHz) is the best 1-core system (BM-5:287).
  • A 2-core IBM Power 550 (4.2 GHz) is the best 2-core system (BM-5:186).
  • A 4-core IBM Power 550 (4.2 GHz) is the best 4-core system (BM-5:154).
  • An 8-core IBM Power 550 (4.2 GHz) is the best 8-core system (BM-5:151).

Source: http://www.ansys.com/services/hardware-support-db.htm (link resides outside of ibm.com)

CD-adapco - STAR-CD Versions 3.20

STAR-CD is a general purpose, unstructured grid, finite volume CFD application. It is efficiently programmed and highly parallel using the MPI communication protocol. STAR-CD is able to solve steady state or time dependent problems with the optional solution of moving grids, thermal transport, chemical reactions, multiphase fluids, and species concentrations. A lower number indicates a faster result.

  • A 1-core IBM System p5 575 (2.2 GHz) is the best 1-core system (A-Class model: 3,830 seconds, Large A-Class model: 35,635 seconds).
  • A 2-core IBM System p5 575 (2.2 GHz) is the best 2-core system (A-Class model: 1,885 seconds, Large A-Class model: 18,807 seconds).
  • A 4-core IBM System p5 575 (2.2 GHz) is the best 4-core system (Engine Block model: 104 seconds, A-Class model: 980 seconds, Large A-Class 9,382 seconds).
  • An 8-core IBM System p5 575 (2.2 GHz) is the best 8-core system (Engine Block model: 56 seconds, A-Class model: 465 seconds, Large A-Class 4,821 seconds).
  • A 16-core IBM System p5 575 cluster (2.2 GHz, equipped with the High Performance Switch Option) is the best 16-core system (A-Class model: 240 seconds, Large A-Class 2,197 seconds).
  • A 32-core IBM System p5 575 cluster (2.2 GHz, equipped with the High Performance Switch Option) produced leadership performance parallel runs for the 32-core A-Class (125 seconds) and Large A-Class (1,107 seconds) test cases.
  • A 48-core IBM System p5 575 cluster (1.9 GHz, equipped with the High Performance Switch Option) produced leadership performance parallel runs for the 48-core A-Class (99.76 seconds) test case.

Source: http://www.cd-adapco.com/products/star_cd/index.html (link resides outside of ibm.com)

MSC.Software

MSC Nastran is a structural analysis program utilizing simple to complex, linear and nonlinear analyses techniques for analyzing stress, vibration, dynamic, nonlinear and heat transfer characteristics of structures and mechanical components.

MSC.Nastran MDR3

  • A 1-core IBM Power 570 (4.7 GHz) is the best 1-core result (xl0tdf1:1632, xl0imf1:1016)
  • A 2-core IBM Power 570 (4.7 GHz) is the best 2-core result (xl0tdf1:863, md0mdf1:469)
  • A 4-core IBM Power 570 (4.7 GHz) is the best 4-core result (xl0tdf1:477)

MSC.Nastran V2005

  • A 1-core IBM Power 550 (4.2 GHz) is the best-in-class 1-core system (BI:5332, OX12:11136, XLTDF:1986, XXAFST:515, XXCMD:5882, XLEMF:1388, XXCMDA:1325)

MD Nastran V2006

  • A 1-core IBM Power 550 (4.2 GHz) is the best 1-core system (XLTDF:2340, XXCMD:5474, XXCMD2:1107, XXIMA:8208)
  • A 2-core IBM System p5-570 (2.2 GHz) is the best 2-core system (XLTDF:1216, XXCMD:3424, XXCMD2:714, XXIMA:1938)
  • A 4-core IBM System p5-570 (2.2 GHz) is the best 4-core system (XLTDF:672, XXCMD:1568, XXCMD2:505, XXIMA:1350)
  • An 8-core IBM System p5-570 (2.2 GHz) is the best 8-core system (XLTDF:424, XXCMD:1071, XXCMD2:411)
  • A 16-core IBM System p5-570 (2.2 GHz) is the best 16-core system (XLTDF:375)

Source: http://www.mscsoftware.com/support/prod_support/nastran/performance/ (link resides outside of ibm.com)

MM5 Weather Model

The PSU/NCAR mesoscale model (known as MM5) is a limited-area, nonhydrostatic, terrain-following sigma-coordinate model designed to simulate or predict mesoscale atmospheric circulation. The model is supported by several pre- and post-processing programs, which are referred to collectively as the MM5 modeling system. A higher number indicates a faster result.

MM5 Version 3.6 (Standard)

  • A 1-core IBM System p5 575 (2.2 GHz) is the best 1-core system (1.62 Gigaflop/sec).
  • A 2-core IBM System p5 575 (2.2 GHz) is the best 2-core system (3.22 Gigaflop/sec).
  • A 4-core IBM System p5 575 (2.2 GHz) is the best 4-core system (6.23 Gigaflop/sec).
  • An 8-core IBM System p5 575 (2.2 GHz) is the best 8-core system (11.89 Gigaflop/sec).
  • A 16-core IBM System p5 575 (2.2 GHz) is the best 8-core system (22.90 Gigaflop/sec).
  • A 32-core IBM System p5 575 (2.2 GHz) is the best 8-core system (41.17 Gigaflop/sec).
  • A 64-core IBM System p5 575 (2.2 GHz) is the best 8-core system (68.56 Gigaflop/sec).

MM5 Version 3.6.2 (Optimized)

  • A 1-core IBM System p5 575 (2.2 GHz) is the best 1-core system (2.33 Gigaflop/sec).
  • A 2-core IBM System p5 575 (2.2 GHz) is the best 2-core system (4.51 Gigaflop/sec).
  • A 4-core IBM System p5 575 (2.2 GHz) is the best 4-core system (8.91 Gigaflop/sec).
  • An 8-core IBM System p5 575 (2.2 GHz) is the best 8-core system (17.05 Gigaflop/sec).
  • A 16-core IBM System p5 575 (2.2 GHz) is the best 8-core system (31.68 Gigaflop/sec).
  • A 32-core IBM System p5 575 (2.2 GHz) is the best 8-core system (55.18 Gigaflop/sec).
  • A 64-core IBM System p5 575 (2.2 GHz) is the best 8-core system (86.20 Gigaflop/sec).

Source: http://www.mmm.ucar.edu/mm5/ (link resides outside of ibm.com)

Back to top

All benchmark claims are based on published information as of October 6, 2011. IBM Power 780 64-core 3.92 GHz and 96-core results submitted as of October 12, 2011.

IBM, the e-business logo, eServer, the eServer logo, AIX, AIX 5L, DB2, DB2 Universal Database, Domino, IBM Virtualization Engine, Micro-Partitioning, OpenPower, Power Architecture, POWER, POWER5, POWER5+, POWER6, POWER6+, POWER7, Power Systems, System p and System p5 are trademarks or registered trademarks of the International Business Machines Corporation in the United States and/or other countries. Other company, product and service names may be trademarks or service marks of others.

UNIX is a registered trademark of The Open Group in the United States and other countries.

SPECjbb, SPECweb, SPECint, SPEComp, and SPECfp are registered trademarks of the Standard Performance Evaluation Corporation (SPEC).

©2011 IBM Corporation. All rights reserved.

We're here to help

Easy ways to get the answers you need.

or call us at
1-866-883-8901
Priority code:
101AR13W

Content navigation

Related links