GPU expansion – the new SciNet ‘Mist’ Cluster

As part of our renewed funding from FedDev Ontario, we are excited to announce the launching of our new contributed GPU-Accelerated Platform in collaboration with SciNet.

This system is a high-performance GPU cluster based on IBM Power System AC922 powered by NVIDIA Tesla V100 GPUs and IBM Power9 CPUs. The systems have NVLink 2.0 technology to accelerate computing by allowing for greater speed, programmability and accessibility of data.

Specifics of the cluster:

  • Mist consists of 54 nodes.
  • Each node has 32 IBM Power9 CPU cores (total core count: 1728, all four-way threaded) and 4 NVIDIA V100 Volta GPUs (total CUDA cores > million).
  • There is 256 GB of RAM per node.
  • EDR Infiniband one-to-one network with GPU-Direct RMDA capability.
  • Shares file systems with the Niagara cluster (parallel filesystem: IBM Spectrum Scale, formerly known as GPFS).
  • No local disks.
  • Theoretical peak performance (“Rpeak”) of 1.6 PF (double precision), 3.2 PF (single precision).
  • Technical documentation can be found at SciNet’s documentation wiki.

The new GPU-Accelerated Platform supports popular AI, Machine Learning and Deep Learning frameworks such as TensorFlow and PyTorch provided as Python Wheels and Conda packages from Watson Machine Learning Community Edition (PowerAI). The platform also has capability of accelerating Molecular Dynamics codes (NAMD, Gromacs, etc) and other computationally intensive applications in computational chemistry, geophysics, next generation sequencing and astronomy.

Further details can be found at SciNet here: https://www.scinethpc.ca/mist

Details on all of our Advanced Computing Platforms can be found here: https://www.soscip.org/platforms