Tango

Tango

Tango is a large AMD Opteron system and is VPAC's fourth and largest cluster. Tango currently has 111 compute nodes, each with two AMD Shanghai 2.5GHz quad core processors for a total of 888 CPUs. In addition to this Tango has a 4 node GPU partition.

Specifications

  • 111 Compute Nodes, 888 CPUs (or more correctly, cores)
    • Each node has 2 AMD Quad Core Opterons™ (2376 (2.3Ghz) or 2379 (2.46Ghz)), 32GB ram and 4 x 320GB disks (1.2TB scratch space)
  • Infiniband Interconnect (~2µs measured MPI latency) on 95 nodes.
  • CentOS 5 Linux

Tango's architecture is similar to Wexstan, but on a much larger scale.

GPU

In addition to this Tango has a 4 node GPU partition

  • The 4 GPU nodes have 1 quadcore processor each with 4GB of RAM and a single 320GB hard drive. The differences are in the type and quantity of GPUs and are as follows:
    • nv1: GeForce GTX 280
    • nv2: GeForce GTX 280 + GeForce GTX 280 (2 gpus)
    • nv3: GeForce GTX 285
    • nv4: Tesla C1060

Compilers

The GNU Compiler Collection (gcc, g++, gfortran) is provided standard as part of the CentOS 5 installation, but we also have include both the Portland Group and Intel compiler suites for better code optimisation.

Recommended Portland Group (PGI) Compiler Flags

The recommended compiler flags for the Portland Group compilers are:

-fastsse -tp k8-64,barcelona-64 -Mipa=fast

To link in the AMD Core Math Libraries add this to your link command:

-l acml

Public Key

The public RSA key fingerprint for tango is c9:2a:cd:04:51:ad:a7:f7:38:bd:a2:fe:a1:5c:c9:79

Featured Media

Selected case studies. See all 6 case studies for Supercomputers.

  • HPC Optimisation Image.jpg
    High Performance Computing Optimisation

    The project’s objective was to develop an optimisation framework for high-performance computing solutions, encompassing a comparative analysis of performance profiling tools with a focus on the appl

  • Aerodynamics modelling
    HPC Cluster for the AutoCRC

    The Cooperative Research Centre for Advanced Automotive Technology (AutoCRC) is a research collaboration aimed at delivering smarter, safer and cleaner manufacturing and vehicle technologies to Austra

  • Nanotube
    Nanoscience and High Performance Computing

    High Performance Computing is used in nanoscience to investigate the relationship between the size, shape and structure of nanoscale materials (nanomorphology) and the thermal and chemical environment