Skip to content

What's New in AAC Bare Metal

This page highlights recent updates and new features for AMD Accelerator Cloud (AAC) Bare Metal clusters.

April 2026 - MI355X GA, ROCm 7.2, Pyxis integration

MI355X cluster is now generally available

The new AMD Instinct MI355X cluster is now generally available on MI355X.amd.com.

Specifications: - 8× AMD Instinct MI355X GPUs per node (288 GB HBM3 per GPU) - 256 CPU cores per node (dual-socket AMD EPYC) - Ubuntu 22.04 OS - High-speed interconnect for multi-node jobs

Access:

ssh <username>@aacXX.amd.com

Partition: 256C8G1H_MI355X_Ubuntu22

See Node Reference Guide for complete specifications and GPU partitioning options.

ROCm 7.2.x promoted as default

ROCm 7.2.0 is now the default and recommended version across both MI325X and MI355X clusters.

Load the module:

module load rocm/7.2.0

ROCm 7.1.3 Release Candidate remains available:
ROCm 7.1.3 RC remains available for compatibility testing or for workloads that specifically require that version. For new workloads, use ROCm 7.2.0.

module load rocm/7.1.3

Legacy versions: ROCm 6.x modules remain available for backward compatibility but are no longer recommended for new workloads.

See AAC Slurm Cluster User Guide for module loading instructions.

Pyxis integrated for containerized workloads

Pyxis, the Slurm plugin for container integration, is now enabled cluster-wide on both MI325X and MI355X.

Benefits: - Run containerized jobs directly with srun --container-image=... - No need to manually run enroot create or enroot start in most cases - Seamless integration with Slurm job scheduler

Example usage:

# Run PyTorch training directly from Docker Hub
srun --container-image=docker://rocm/pytorch-training:v25.5 \
  --container-mounts=$HOME:/workdir \
  --container-workdir=/workdir \
  python train.py

# Use local Enroot .sqsh image
srun --container-image=/shared/data/my-image.sqsh \
  --container-writable \
  bash -c "python train.py"

See Using Enroot with Pyxis for more examples.

Anaconda 25.5.1 module available on MI325X

Anaconda3 25.5.1 is now available as an environment module on MI325X (MI325X cluster).

Load the module:

module load anaconda3/25.5.1

Create and activate environments:

conda create -n myenv python=3.11
conda activate myenv

Note: Anaconda is currently available on MI325X only. If you need it on MI355X, contact cluster operations or use containerized Python environments.

See Using Anaconda on AAC for complete usage guide.

New documentation pages

Several new documentation pages have been added to help you get the most out of AAC clusters:

Previous updates

For Plexus-specific release notes, see Plexus What's New.

Feedback and support

If you encounter issues or have questions: - Contact your AMD sponsor - Submit a support ticket through the AAC portal - Check the AAC Slurm Cluster User Guide for troubleshooting tips