Skip to content

GPU Development Platform

As part of the DANCE infrastructure, the GPU cloud computing service at CPPM offers a convenient environment for users who need GPU acceleration for their developments, training, or testing.

What is it?

The GPU development platform at CPPM is a service based on the cloud computing platform. It allows users to run GPU-accelerated applications on virtual machines (VMs) with GPU passthrough.

This platform is not intended for production workloads but is ideal for:

  • Code development and testing
  • Training on GPU tools and workflows
  • Educational and proof-of-concept projects

Hardware

The infrastructure currently consists of two compute nodes with the following specifications:

  • 2× Intel Xeon Silver 4210 CPUs (20 cores total per node)
  • 128 GB RAM
  • 1 TB SSD
  • 3× NVIDIA RTX2080Ti GPUs per node
  • 100 Gb/s Ethernet connection

Access Options

Users can access the GPUs through different types of virtual machines:

  • Multi-user GPU Server
  • Shared VM with 3 GPUs
  • No cloud experience required

  • Direct access with a CPPM account

  • Single-user GPU Server

  • Dedicated GPU per user
  • Requires knowledge of OpenStack
  • Requires a CPPM OpenStack account

⚠️ The single-user VMs were underutilized, so we now also provide a second multi-user GPU server that hosts a JupyterHub environment for interactive work.

How to Get Access

Access is available to all CPPM members.

  • For Multi-user GPU servers, you only need a CPPM account.
  • For Single-user GPU servers, request an OpenStack account.

To request access or report issues, open a GitLab issue on our GPU documentation repository:
👉 GitLab Issues – GPU Cloud

Documentation

For more information on how to use the GPU cloud service, refer to the documentation.

For Production GPU Needs

For large-scale production GPU computing, consider using:

We can assist you in applying to these platforms—just open an issue on GitLab.