Hosted at the University of Mons, this cluster is made of 14 computing nodes, each with two AMD Zen5 9455 48-Core Processors at 3.1 GHz, 768 GB of memory, and 7 TB NVMe SSD local scratch directory.
The compute nodes are interconnected with a 25 Gigabit Ethernet network.

Suitable for:

Long (max. 21 days) shared-memory parallel jobs (OpenMP or Pthreads), or resource-intensive (CPU speed and memory) sequential jobs.

Resources

  • Home directory (80GB quota per user)
  • Local working directory $LOCALSCRATCH - 7TB NVMe SSD
  • Global working directory $GLOBALSCRATCH - 98TB total
  • CÉCI Common Storage with $CECIHOME and project folders
  • No internet access from nodes
  • Queues* :
    • Batch queue: Max 5 days
    • Long queue: Max 21 days
    • Current core limit: 72 cores/user

Software Stack

The most complete software stack can be found within the default 2025b release, using GCC 14.3 for full native support of the AMD Zen5 processor.

Before installing your own software, check whether it is already available on Dragon3 with the module spider <software_name> command.

If you need Python, we highly recommend using the pre-installed Python 3.13.5 (module load Python/3.13.5-GCCcore-14.3.0) along with its optimized bundles such as SciPy-bundle, Matplotlib, ASE, scikit-learn, and scikit-image.

Access / Support

SSH to dragon3.umons.ac.be (port 22) via your university gateway, with proper login and id_rsa.ceci file.

Support : CECI Support form

Server SSH key fingerprints:

What’s this ?

ECDSA:

SHA256:iGnsFs7hLK1O6lP0IqMpUcZFby3xeay7CZghZdev6yo

ED25519:

SHA256:6qfUCyW7qS4K/PiFqwU2C/GtGdp7iHXD0+joQgSAgEo

RSA:

SHA256:AEdYjw8Ym05vyLPMbDwXhlfDF/Uz286KthIaiP/Uxag

cluster image 1