NCCS GPU System Details

System Architecture

This system was designed to be quickly used by Discover users and is coupled to the Discover IBM GPFS file system. Therefore, the operating environment is consistent with the Discover operating system, software stack, tools, and file system. The Portland Group Inc. (PGI) CUDA compiler with FORTRAN support is available for use on these nodes.  Detail on how to run CUDA compiler will be available soon.

Overall Specifications

  • Manufacturer: IBM
  • 528 Total Cores (General Purpose Intel Cores)
  • 95,040 Total Cores (GPU Streaming Cores)
  • Production: 3Q 2013

33 Compute Nodes Each Configured with the Following

  • Dual-socket, 8-core 2.6 GHz Intel Xeon Sandy Bridge processors(16 processors per node)
  • 64 GB of memory per node(4GB/core)
  • Interconnect: Infiniband FDR
  • NVIDIA K40 GPUs (one per node)
  • Operating System: SLES11 SP3

33 Graphical Processing Units each configured with the following:

  • 1 Tesla GPU K40M (GK110B)
  • 2,880 CUDA cores
  • ECC Memory (internal and external)
  • 12 GB of GDDR5 memory
  • 1.43 Tflops of double precision floating point performance (peak)
  • 4.29 Tflops of single precision floating point performance (peak)
  • 288 GB/sec memory bandwidth
More information on Tesla k40 GPU Accelerators