PetaFLOP AI Performance
Up to 1 PFLOP of FP4 compute for training and inference workloads.
FACILITIES
Facilities
The Centre for AI is equipped with advanced computational and research facilities to support high-end AI experimentation, model training, and applied research.
Computing infrastructure
NVIDIA DGX Spark™
The most recent addition to our computing infrastructure, the NVIDIA DGX Spark delivers up to one petaFLOP of FP4 AI performance in a compact, energy-efficient form factor. It enables faculty and students to run large-scale AI models and conduct cutting-edge research directly on campus.
Key System Highlights
Energy-Efficient Design
Desktop-class form factor with data-center-grade performance for sustainable research.
GB10 Superchip
Powered by the NVIDIA Blackwell GPU and 20-core ARM CPU with 128 GB unified memory.
Seamless Integration
Fully compatible with existing HPC workflows, Docker containers, and Jupyter environments.
Research laboratory
High Performance Computing Facility Overview
Server infrastructure
HPC Server Details
Vega 1 (Server 1)
Processor
2 x Intel Xeon Gold 5220 2.2G, 18C/36T, 24.75M Cache
Graphics Card
2 x NVIDIA Tesla V100 16G Passive GPU
Hard Drives
4 x 1.2TB 10K RPM SAS 12Gbps
Memory
6 x 32GB RDIMM, 3200MT/s, Dual Rank
OS
Ubuntu 18.04.05 LTS
Applications
Docker (Containers: TensorFlow, MXNet, PyTorch)
Vega 2 (Server 2)
Processor
2 x Intel Xeon Gold 5220 2.2G, 18C/36T, 24.75M Cache
Graphics Card
2 x NVIDIA Tesla V100 16G Passive GPU
Hard Drives
4 x 1.2TB 10K RPM SAS 12Gbps
Memory
6 x 32GB RDIMM, 3200MT/s, Dual Rank
OS
Ubuntu 21.04 LTS (GUI Enabled)
Applications
Keras, TensorFlow, Jupyter Notebook
Lab workstations
Desktop Lab & Developer Kits
Hardware Inventory
Desktops
20 x OptiPlex 3070 SFF — Processor: Intel Core i7-9700 (8 cores), Memory: 16GB DDR4
Developer Kits
Jetson Nano (2), Jetson TX2 (2)