High Performance Computing (HPC) Facility

High Performance Computing Facility (HPC)

Most of the modern AI/ML algorithms, especially deep learning techniques, often require exceptional amounts of computation and data for achieving acceptable performance in real world applications. Other classical systems that employ beamforming, tracking, localisation, noise control or equalization also give rise to problems of high computational cost. High-performance computing (HPC) is the use of distributed and parallel processing techniques for solving such complex computational problems. Recognising the computational needs in various research frontiers such as image processing, signal processing and machine learning, the department implemented an HPC facility with financial assistance from the university, DST and RUSA.

Ongoing Projects
  • Machine Learning Models for Underwater Image Enhancement and Content Analysis
  • Development of Biomimetic Approaches for Sonar Systems
HPC Specifications
  • Number of Compute Nodes : 6 (4 GPUs/Node)
  • Number of Master Nodes : 3
  • Total number of GPUs : 24
  • Total number of CUDA Cores : 96256
  • Number of Administrative/Master Nodes : 3
  • Aggregate GPU RAM : 284 GB
HPC Specifications
  • Aggregate Single Precision Performance : ~250 TFLOPS
  • Intel Broadwell E5 2650 V4 Processor 12 Cores per Socket
  • Total number of CPU Cores: 144 Cores
  • Aggregate CPU TFLOPS: 6 TFLOPS
  • Aggregate RAM : 1536 GB
  • Total Storage : ~ 96 TB