New GPU Nodes with NVIDIA H200 Available

Published: 28 Aug 2025 by NEMO Team

We are happy to announce that two new GPU nodes equipped with the latest NVIDIA H200 GPUs are now available on NEMO2.

Each node is configured with:

8× NVIDIA H200 GPUs
192 CPU cores
1.5 TB of RAM
3.8 TB local NVMe storage

One of these nodes has already been integrated into NEMO2 and is ready for testing.
To use it, please submit your jobs with the partition h200.

Please note: no software modules are installed yet on these nodes.
You may need to bring your own software environment (e.g. via containers or conda) for testing.

Example job script

#!/bin/bash
#SBATCH --job-name=test-h200
#SBATCH --partition=h200
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=8
#SBATCH --mem=64G
#SBATCH --time=02:00:00

# load CUDA if available, otherwise use your own environment
module load system/cuda/12.6.0 || true

srun nvidia-smi

An SSH configuration issue on the login nodes has been fixed. SSH keys stored in local authorized_keys files are no longer accepted. This behavior was never intended - please use SSH keys configured in login.bwidm.de instead.

Over the holiday period, we’re increasing queue limits to 10,000 jobs while reducing cluster capacity for energy savings. Normal operation resumes automatically on January 7, 2026.

NEMO2 has undergone a rolling upgrade with security, firmware/BIOS, and Slurm 25.05.5 updates. We planned and executed all steps to keep downtime for users as minimal as possible. Running jobs continued without interruption throughout the process.