NCSA Home
Contact Us | Intranet | Search

Transitioning from the NCSA IBM pSeries 690 (copper)

  • NCSA IBM pSeries 690 Retirement Notice

  • Updates:

    • The cap queue will no longer accept job submissions starting at noon CT September 14.
    • Jobs that the scheduler calculates will not finish before midnight on September 30th won't be started. This means that as September 30 approaches, submitted jobs have a lesser chance of starting. Based on the current job load, we expect that jobs submitted after September 26 may never get an opportunity to start.
    • As of Sept 6 2007, in preparation for copper's impending retirement the wall clock limit on the cap queue has been reduced to 300 hours.
    • You can save your copper home directory to a tar file named copperhome.tar in your UniTree account by executing the following command:
           cu12% cd $HOME; msscmd "tar cf copperhome.tar ."
      

  • NCSA HPC Systems Summary Information:

    Category Cobalt (*) Tungsten Mercury Abe (**)
    Machine Type IA-64 Itanium SGI Altix (SMP) Dual processor Xeon IA-32 Cluster Dual processor IA-64 Itanium Cluster Dual socket quad core Intel 64 Cluster
    Machine Page Link Link Link Link
    Programming Model OpenMP, MPI MPI MPI MPI, MPI/OpenMP hybrid
    Compilers Intel
    byte order for Fortran unformatted files Little-endian (see here for further information)
    Batch System PBSpro LSF PBS Torque
    Max Wall Clock Time 18 hours
    (8 CPUs/144 hours extended queue will be available in the future)
    100 hours 96 hours 8 hours
    Maximum Memory per node/
    Maximum CPUs per node
    3 TB/512 CPUs 3 GB/2 CPUs 12 GB/2 CPUs 8 GB/8 CPUs (cores)
    Vendor Math Library SGI SCSL, Intel MKL Intel MKL
    SU Conversion Factor from copper 1/3.2 1/2.0 1/2.2 1/3.5
    (*) Primarily intended for applications requiring large shared-memory and high levels of parallelism. Tungsten and Mercury are recommended for applications using small number of processors and do not require shared-memory.
    (**) Capability resource - production jobs should typically use at least 1000 cores

  • Applications Software on NCSA HPC Systems

  • TeraGrid Resources Catalog (for resource availability across the TeraGrid)

Please contact the NCSA HPC Consulting Office at 217-244-1144 or by e-mail at consult@ncsa.uiuc.edu with any questions or problems.