Handson Supercomputing

This lecture introduces to the practical aspects of using fast workstations and a real supercomputer.

working on a fast workstation

The first workstation is pascal.math.uic.edu, a Microway numbersmasher Xeon + Tesla GPU server (2016):

  • two 22-core Intel Xeon E5-2699v4 Broadwell at 2.20GHz,

  • 256GB of internal memory at 2400MHz,

  • NVIDIA Tesla P100 16GB Pascal GPU accelerators, 4.7 TFLOPS (FP64) peak performance.

The newer ampere.math.uic.edu is a Microway 2U Xeon + NVIDIA GPU server (2024):

  • two 24-core Intel Xeon 5318Y Ice Lake-SP, up to 3.40GHz,

  • 256GB of internal memory at 3200MHz,

  • NVIDIA Ampere A100 80GB GPU accelerator, 8.6 TFLOPS (FP64) peak performance. TensorCore performance: up to 19.5 TFLOPS (FP64).

Login via ssh at a Terminal or PowerShell window. To transfer files use secure copy scp.

using a real supercomputer

Access has been granted through https://access-ci.org.

The workflow typically includes the following:

  1. request access to the cluster

  2. login to the cluster

  3. user work spaces and directories

  4. requesting and running software

  5. submitting jobs

  6. monitoring a job