National Computational Infrastructure
NCI National Facility
SGI Altix AC Hardware Details
Image of Alphaserver SC

One of the three rows of the Alphaserver SC system with the robotic tape mass data storage system in the background

APAC system Specifications of the peak system:
  • 126 nodes each containing:
    • 4 x 1GHz ev68 (Alpha 21264C) cpus
    • L1 cache (on chip): 64 kbytes (I) + 64 Kbytes (D)
    • L2 cache (off chip): 8 Mbytes per cpu
    • between 4 and 16GB of RAM (currently 36 8GB nodes and 4 16GB nodes)
    • between 2 and 6 36GB SCSI disks
  • Aggregate Specfprate_base (of the compute nodes only) 4050
    Peak theoretical performance of over 1Tflop and achieving ~820 Gflops on Linpack for over 80% performance achievable
  • Total of ~700GB of RAM
  • 2.88 TB global user storage (additional to node disks)
  • 576 GB system storage
  • Total of 13.1 TB of disk
  • Leading edge Quadrics "Elan3" interconnect:
    • fat-tree topology with "direct memory access" and hardware broadcasts (single rail)
    • MPI latency of < 5 usecs
    • MPI bandwidth of 250 Mbytes/sec bidirectional
The GrangeNet broadband network is used to connect to institutions and centres that have signed up as grangenet partners. This allows fast access (gigabit) to the system for distributed projects, especially those involving visualisation or remote data repositories. The status of the GrangeNet connections is monitored via the GrangeNet weather station.
AlphaServer SC Systems General Documentation on Alphaserver systems on HP's web site.
Email problems, suggestions, questions to