BM Seer Unofficial thoughts from an anonymous Sun employee

ABAQUS V6.7 Benchmarks Sun Blade X6250 Cluster World Record

Friday Jan 25, 2008

The ABAQUS "Explicit" benchmark test suite was run on a mini cluster of Sun Blade X6250 blades with the recently announced 3.33 GHz dual-core Intel 5260. The Sun Blade X6250 mini cluster beats all posted results at the ABAQUS V6.7 website up to the eight cores.

  • The closest posted results from a competitor's platform were primarily from an HP XC with dual-core 3GHz 5160 processors and to a limited degree (at the 4 "cpu" level) by an Intel Supermicro with 3GHz quad-core E5472's.
  • In runs of the six cases in the benchmark test suite, the X6250 cluster was nominally 17% faster than the best results coming either from the top HP or Intel cluster over the 4-core levels considered and considering results for all 6 test cases.
  • The scalability efficiency of the X6250 cluster ranged from 100% (at 1 core) to 81% (geometric mean at 8 cores) and considering all 6 test cases at each of the four core levels.

    • Four 2 socket Sun X6250 blades with Infiniband interconnects were used and runs were made at different core levels: 1, 2, 4, and 8. Comparisons are presented against the current leading competitors' results also obtained with high performance interconnects and posted at the ABAQUS V6.7 website. This includes results from IBM, HP, and Intel platforms and clusters with current dual-core and quad-core Intel processors.

    ABAQUS V6.7 "Explicit" Benchmark Test Suite, time in elapsed seconds

    Please note, this table has been modified since the original posting to correct the table and make sure only V6.7 results are shown, sorry for the confusion, but the Sun internal information sites changed since my posting.
    System CPU Benchmark Test
    e1 e2 e3 e4 e5 e6
    One core results
    Sun Blade X6250 3.33GHz DC 5260 23565 12399 11037 4884 4648 11975
    Sun Blade X6250 3.0GHz QC 5365 26401 14236 12302 5456 5349 13266
    Intel Supermicro 3.0GHz QC E5472 24815 13738 12504 5273 5299 13456
    HP XC 3.0GHz DC 5160 23957 13659 11289 5157 5122 12601
    Bull R440 3.0GHz DC 5160 25132 14086 12237 5352 5231 13213
    Two core results
    Sun Blade X6250 3.33GHz DC 5260 12008 6465 5218 2647 2447 6739
    Sun Blade X6250 3.0GHz QC 5365 14262 7501 6379 2959 2742 7486
    Intel Supermicro 3.0GHz QC E5472 14060 7151 6341 2900 2693 7880
    HP XC 3.0GHz DC 5160 13229 6998 6201 2838 2657 7336
    Bull R440 3.0GHz DC 5160 13859 7283 6575 2997 2756 7752
    Four core results
    Sun Blade X6250 3.33GHz DC 5260 7868 3888 3064 1482 1328 4025
    Sun Blade X6250 3.0GHz QC 5365 8595 4195 3372 1577 1440 4375
    Intel Supermicro 3.0GHz QC E5472 8264 3857 3438 1616 1440 4534
    HP XC 3.0GHz DC 5160 9843 4434 4413 1856 1619 5235
    Bull R440 3.0GHz DC 5160 10067 4559 4485 1964 1651 5378
    Eight core results
    Sun Blade X6250 3.33GHz DC 5260 5209 2439 1922 979 736 2510
    Sun Blade X6250 3.0GHz QC 5365 5650 2556 2158 1090 824 2774
    Intel Supermicro 3.0GHz QC E5472 6077 2473 2529 1205 910 3339
    HP XC 3.0GHz DC 5160 5140 2311 2280 1074 823 2948
    Bull R440 3.0GHz DC 5160 5366 2406 2303 1127 860 3092

    About The ABAQUS Explicit Module

    This module designed for crash and high velocity impact analyses is very scalable and analysis models tend to be very large similar to CFD models. Timely results are best obtained using multiple processing units for typically large jobs either on a single multi core server in smp mode or on a multi node cluster of multi core platforms interconnected in dmp mode.

    • The test cases in the ABAQUS "Explicit" benchmark test suite do not require much memory (all around a few hundred megabytes)
    • The ABAQUS test cases scale very well up to 16 cores. All of the solvers in the Explicit module work in dmp mode on clusters. The ABAQUS default mode for MPI is HP-MPI.
    • Based on the maximum physical memory on a platform the user can stipulate the maximum portion of this memory that can be allocated to the ABAQUS job. This is done in the "abaqus_v6.env" file that either resides in the subdirectory from where the job was launched or in the abaqus "site" subdirectory under the home installation directory.
    • The test cases for the ABAQUS benchmark test suites all have a substantial I/O component. This I/O activity is primarily associated with temporary scratch files. Performance will be enhanced by using the fastest available drives and striping together more than one of them or using a high performance disk storage system with high performance interconnects.

    System Configuration

  • 4 Sun Blade X6250
  • 3.33 GHz dual-core Intel 5260
  • 2 internal striped 15K SAS drives (cluster shared file system)
  • Infiniband (Voltaire) interconnects

  • 64-bit SUSE Linux Enterprise Server SLES 10
  • Voltaire OFED GridStack-4.1.5_7-sles-k2.6.16.21-0.8-smp-x86_64
  • HP-MPI
  • ABAQUS V6.7 Explicit Module
  • ABAQUS 6.7 Explicit Benchmark Test Suite

    • Disclosure Statement:

    The following are trademarks or registered trademarks of Dassault Systems or its subsidiaries in the United States and/or other countries: Abaqus, Abaqus/Standard, Abaqus/Explicit. All information on the ABAQUS website is Copyrighted 2004-2007 by Dassault Systemes. Results from http://www.simulia.com/support/v67/v67_performance.html as of Jan. 18, 2008.

    [2] Comments
    Like this post? del.icio.us | furl | slashdot | technorati | digg
    Posted at 06:07AM Jan 25, 2008 in Benchmarks
    Tags:
    Comments:

    The CPU entries for the single-core HP XW 6200 and IBM X336 seem to be missing the processor model information. Every other entry in your table seems to have it. Also, given your digs at others in:

    http://blogs.sun.com/bmseer/entry/hp_dl580_g5_4_qc

    "It is so easy to measure you wonder why some vendors hide their wattages."

    Where are the power figures for the x6250?-)

    Posted by rick jones on January 28, 2008 at 08:09 AM PST #

    You are right, if I was King those benchmark people would all do it.

    I'll dig around and see if those guys have that power data and post it.
    Maybe I'll even set up a rig myself and measure it.

    You should do the same!

    Posted by BM Seer on January 29, 2008 at 08:21 AM PST #

    Post a Comment:
    Comments are closed for this entry.