Skip to content

FF chevron

Deepthought2 Cluster

The Deepthought2 HPC cluster was the flagship cluster at the University of Maryland from when it came online in July 2014 until just before it was retired in September 2022. The cluster is no longer available for use. It consisted of:

  • 444 compute nodes: each with: - Dual IvyBridge E5-2680v2 CPUs (2.8 GHz), with 20 cores per node - 128 GB of RAM per node (6.4 GB/core) (DDR3 at 1866 MHz) - 750 GB of node local /tmp space per node - FDR Infiniband
    • 40 compute nodes: each with:
      • Dual IvyBridge E5-2680v2 CPUs (2.8 GHz), with 20 cores per node
      • 128 GB of RAM per node (6.4 GB/core) (DDR3 at 1866 MHz)
      • 750 GB of node local /tmp space per node
      • FDR Infiniband
      • Two NVIDIA Tesla K20m (GK110GL) GPUs (CUDA compute capability 3.5)
    • 4 Large memory nodes each with:
      • Quad IvyBridge E5-4640v2 CPUs (2.2 GHz), with 40 cores per node
      • 1024 GB of RAM per node (25.6 GB/core) (DDR3 at 1333 MHz)
      • 750 GB of node local /tmp space per node
      • FDR Infiniband
    • 1 PB scratch storage (lustre)
Description Number of nodes Processor Cores/node Mem/node (GiB) Mem/core /tmp size (TB) GPUs/node
Standard compute 444 Intel Ivybridge 20 128 6.4 0.75 none
Large memory 4 Intel IvyBridge 40 1024 25.6 0.75 none
GPU 40 Intel Ivybridge 20 128 6.4 0.75 Two K20m

The standard compute and GPU nodes had dual Intel E5-2680v2 IvyBridge processors with 10 cores per CPU and a base speed of 2.8 GHz.

The large memory nodes had quad Intel E5-4640v2 IvyBridge processors with 10 cores per CPU and a base speed of 2.2 GHz.

The GPU nodes had dual NVIDIA K20m (GK110GL) GPUs with 5 GB of GPU RAM (using the Kepler architecture supporting CUDA compute capability 3.5).

Note

The Deepthought2 cluster was retired in 2022. It is no longer available for computation. The information in this page is for historical records only.