Computing Resources

We provide computing resources to develop AI applications

Computing Resources

We provide computing resources to develop AI applications

AI Services Computing Resources

The AI Service Centre provides a heterogeneous AI infrastructure that meets the requirements and needs of research, startups, or small/medium-sized enterprises. In addition to a specialized training and inference cluster, it includes an ARM-based system and small and medium general-purpose servers. This infrastructure enables:

  • Exploration of edge applications
  • Comparison of costs and performance of different hardware types
  • Flexible solutions in the development, optimization, and implementation of complex AI models in production environments

Computing Infrastructure

Assisted Requirements

  • GPU-based systems for training and inference
  • High-performance computing (HPC) resources
  • Cost and performance analysis for various workloads and hardware
  • Please have a look at our portal for submitting a hardware request.

Heterogeneous Infrastructure

  • Availability: Starting at the end of 2023
  • Target Users: Research, startups, small/medium-sized enterprises
  • Components:
    • Specialized training and inference cluster
    • ARM-based system
    • Small and medium general-purpose servers
  • Features:
    • Edge application exploration
    • Hardware cost and performance comparison
    • AI model development, optimization, and production implementation

AI Infrastructure Details

Training Cluster

  • Hardware: NVIDIA DGX-H100 base pods (80 GB VRAM)
  • Connectivity: 400 Gb/s Infiniband and/or 200 Gb/s Ethernet
  • Features: Fast memory system with GPU to persistent memory direct writes

Inference Cluster

  • Hardware: NVIDIA A30 tensor-core GPUs
  • Connectivity: 25 Gb/s Infiniband and/or 40 Gb/s Ethernet
  • Purpose: Applying trained models to new data inputs

Overview

  • Start Date: End of 2023
  • Offerings: Curated data, models, software stacks
  • Services: Consulting for model, software stack, and hardware selection
  • Consulting: Guidance on cloud, on-premise, and hybrid solutions

Procedure: Accessing Resources

  1. Application: Submit a project outline via the application form
  2. Confirmation: Receive a commitment based on resource availability
  3. Terms: Review and establish conditions of use
  4. Access: Granted to computing resources
  5. Documentation: Recording of results and artifacts