Clusters at a Glance

NYU Greene Supercomputer

The Greene cluster is named after Greene Street in SoHo, a neighborhood in Lower Manhattan near NYU. The cluster has also "green" characteristics, such as most of the cluster nodes are water cooled and it is deployed in a power efficient data center.

Highlights Hardware
  • Greene is the primary NYU HPC cluster
  • General purpose HPC cluster, suitable for the majority of computing and data analytics tasks, such numerical simulations and AI.
  • Measured performance (from CPUs only) using LinPack: 2.088 PF
  • Ranked #271 on the June 2020 edition of the top500 list (based on an extrapolated performance of 1.731 PF)
  • Using SLURM for job submission
  • Number of nodes: 665
  • CPU cores:  31,584 Intel cores
  • NVIDIA GPUs:  332 (V100 and RTX8000)
  • Total RAM: 163 TB
  • Total Disk space:  9 PB 
  • GPFS file systems
  • HDR Infiniband interconnect

Peel (Hadoop)

PEEL (פיל) is Currently under construction. It will become the main NYU Hadoop cluster during the spring 2020 semester.

Hudson HPC/AI Cluster

Highlights Hardware
  • Hudson is Currently under construction. 
  • This cluster is suitable for AI and workloads utilizing GPU accelerators
  • Theoretical performance: 1 PF (single precision)
  • Jobs are submitted using SLURM
  • 20 nodes each equipped with 8 GPUs
  • AMD GPUs: 160 MI-50
  • CPU Cores: 960 EPYC Rome 
  • Total RAM: 10 TB
  • HDR Infiniband interconnect 
  • Same filesystems as Greene cluster