HPC2002 Result: Hewlett-Packard Company HP Cluster Platform 4000 w/XC (DL145 G2)

Benchmark	Reference Time	Runtime	Ratio
361.wrf_m	86400	112	769
SPECenvM2002	769

Benchmark

Reference
Time

Runtime

Ratio

361.wrf_m

86400

112

769

SPECenvM2002

769

Hardware

Hardware Vendor:

Hewlett-Packard Company

Model Name:

HP Cluster Platform 4000 w/XC (DL145 G2)

CPU:

AMD Opteron(tm) Processor 252

CPU MHz:

2600

FPU:

Integrated

CPU(s) enabled:

128 cores, 128 chips, 1 core/chip

CPU(s) orderable:

1 to 2 per node

Primary Cache:

64KBI + 64KBD (on chip) per core

Secondary Cache:

1 MB on chip

L3 Cache:

Other Cache:

None

Memory:

4 GB DDR PC3200 per node (8x512K)

Disk Subsystem:

1x80GB SATA disk (root)

Other Hardware:

See below for a more complete system descrition

Software

Parallel:

MPI

Processes-Threads:

128

MPI Processes:

128

OpenMP Threads:

Operating System:

XC Linux for High Performance Computing v3.0

Compiler:

Pathscale 2.2 Fortran Compiler
Pathscale 2.2 C Compiler

File System:

NFS Shared File System

System State:

Multi-user

Other Software:

HP-MPI 2.1.1, LSF 6.1.7, SLURM 0.5.0-10

Notes / Tuning Information
Peak Flags: mpif90 -Ofast -I. -I${NETCDF}/include FPORTABILITY = -DF2CSTYLE mpicc -Ofast -I. -I${NETCDF}/include CPORTABILITY = -DSPEC_HPG_MPI2 Preprocessing: -I. -traditional Link Flags EXTRA_LIBS = -L${NETCDF}/lib -lnetcdf ENV_SPEC_HPG_PARALLEL=MPI Flags file description HP-MPI-Pathscale-20050913.txt Alternate Source used for Peak: hplinux fix errno.h include file problem under #ifdef T3D Available as SPEC HPC2002 Source: env2002-src_hp-errno-fix-20050907.tar.gz Peak User Environment: bsub -n 128 ./runspec -c linux_amd_psc --reportable env_m use_submit_for_speed=1 submit = \$MPI_ROOT/bin/mpirun -srun taskset 0x3 $command < /dev/null -srun launches one process per procesor consisting of the command which follows. taskset execs its arguments and bind proceses to processors according to mask. $command is generated by the SPEC tools. System Description HPC Cluster Platform is a Hewlett-Packard preconfigured and factory built hardware and software solution scalable from 5 to 512 nodes. The product used in these submissions is an HP Cluster Platform model 4000. For Product Information see www.hp.com and search for HPC Clusters Platforms http://www.hp.com/techservers/clusters/ucp/index.html XC Clusters http://www.hp.com/techservers/clusters/xc_clusters.html For detailed quick specs, search www.hp.com and search for: HP Cluster Platform 3000 and HP Cluster Platform 4000 http://h18000.www1.hp.com/products/quickspecs/12306_div/12306_div.HTML XC System Software V2.1 quickspecs http://h18000.www1.hp.com/products/quickspecs/12094_div/12094_div.HTML Underlying Cluster compute nodes: HP ProLiant DL145 G2 server 64 compute nodes used for this run. Network (for computation) Voltaire Infiniband HCA 400Ex Voltaire leaf switches - ISR 9024 (1 per 12 nodes) Voltaire aggregation switch - ISR 9288 (12 ports per leaf switch) Network (for File Server ) ProCurve 2848 Gb Ethernet Switches (1 per 40 nodes) File Server HP Proliant DL585 two AMD Opteron (tm) Processor 850 2400MHz 8 GB Memory 4 2GB PC2100 Dimms 146 GB SCSI 10000 Disk Additional Linux Software Netcdf 3.5.1 source obtained from http://www.unidata.ucar.edu/packages/netcdf/ build with (./configure ; make )

Notes / Tuning Information

 Peak Flags:
   mpif90  -Ofast -I. -I${NETCDF}/include 
     FPORTABILITY =  -DF2CSTYLE

   mpicc -Ofast -I.  -I${NETCDF}/include
     CPORTABILITY = -DSPEC_HPG_MPI2

   Preprocessing:
     -I. -traditional

   Link Flags EXTRA_LIBS = -L${NETCDF}/lib -lnetcdf

   ENV_SPEC_HPG_PARALLEL=MPI

  Flags file description HP-MPI-Pathscale-20050913.txt 

 Alternate Source used for Peak:
   hplinux 
      fix errno.h include file problem under #ifdef T3D
         Available as SPEC HPC2002 Source: 
           env2002-src_hp-errno-fix-20050907.tar.gz

 Peak User Environment:
   bsub -n 128 ./runspec -c linux_amd_psc --reportable env_m

   use_submit_for_speed=1
   submit = \$MPI_ROOT/bin/mpirun -srun 
        taskset 0x3 $command < /dev/null

      -srun launches one process per procesor
         consisting of the command which follows.
        taskset execs its arguments and bind proceses
        to processors according to mask. 
         $command is generated by the SPEC tools.

 System Description
   HPC Cluster Platform is a Hewlett-Packard preconfigured 
   and factory built hardware and software solution scalable 
   from 5 to 512 nodes. The product used in these submissions 
   is an HP Cluster Platform model 4000.

   For Product Information see www.hp.com and search for
     HPC Clusters Platforms
       http://www.hp.com/techservers/clusters/ucp/index.html
     XC Clusters 
       http://www.hp.com/techservers/clusters/xc_clusters.html

   For detailed quick specs, search www.hp.com and search for:
     HP Cluster Platform 3000 and HP Cluster Platform 4000 
       http://h18000.www1.hp.com/products/quickspecs/12306_div/12306_div.HTML
     XC System Software V2.1 quickspecs   
       http://h18000.www1.hp.com/products/quickspecs/12094_div/12094_div.HTML

 Underlying Cluster compute nodes:
   HP ProLiant DL145 G2 server 
   64 compute nodes used for this run.

 Network (for computation)
   Voltaire Infiniband HCA 400Ex
   Voltaire leaf switches - ISR 9024 (1 per 12 nodes)
   Voltaire aggregation switch - ISR 9288 (12 ports per leaf switch)

 Network (for File Server )
   ProCurve 2848 Gb Ethernet Switches (1 per 40 nodes)

 File Server 
   HP Proliant DL585 
      two AMD Opteron (tm) Processor 850 2400MHz
      8 GB Memory 4 2GB PC2100 Dimms 
      146 GB SCSI 10000 Disk

 Additional Linux Software
   Netcdf 3.5.1 source obtained from 
     http://www.unidata.ucar.edu/packages/netcdf/
       build with (./configure ; make )

First published at SPEC.org on 28-Sep-2005

Generated on Wed Sep 28 11:52:54 2005 by SPEC HPC2002 HTML formatter v1.01