Computing Resources

HPCL Overview

The High-Performance Computing Lab (HPCL) provides users with a powerful cluster of 34 high-performance computer nodes.

Individual Node Specifications

Each node is equipped with the following resources:

• Processor (CPU): 13th Gen Intel(R) Core(TM) i9-13900KF
  • Cores: 24 per node
  • Threads: 32 per node
• Graphics Processing Unit (GPU): NVIDIA GeForce RTX 4080
  • CUDA Cores: 9,728 per node
• Memory (RAM): 32 GB DDR5 4800MHz
• Local Storage: 512GB NVMe SSD (for scratch space)
• Shared Storage: Access to a shared /home directory.

Cluster-Wide Specifications

Collectively, the HPCL cluster offers the following computational resources:

• Total Physical Cores: 816
• Total Threads: 1088
• Total CUDA Cores: 330,752
• Total RAM: 1088 GB DDR5
• Total Local Scratch Storage: Approximately 17.4 TB

Supported Capabilities

Beyond the raw hardware, the HPCL cluster and its management software support:

• Message Passing Interface (MPI): Primarily designed for running MPI-aware parallel applications.
• Diverse Application Needs: Capable of supporting applications requiring license servers, alternative network configurations, and batch processing.
• Experimental Hadoop Support: Utilizing the Magpie framework for deploying Hadoop within the cluster environment.
• Ollama: Support for running and deploying large language models locally using Ollama, leveraging the available CPU and GPU resources on the nodes.
• R and R Studio: Comprehensive support for statistical computing and analysis using the R programming language and the R Studio integrated development environment.