Voyager enters ACCESS allocations Phase

Voyager Supercomputer Enters ACCESS Allocations Phase

As of June 1, 2025, SDSC’s Voyager supercomputer has transitioned from its testbed phase to the ACCESS Allocations Phase. All research use now requires an official ACCESS allocation.

About Voyager

Voyager is NSF’s first AI-focused supercomputer, built around Habana Gaudi training and Goya inference processors. It offers:

• 42 Supermicro X12 Gaudi AI Training Systems with 336 Habana Gaudi processors.
• 16 Habana Goya processors for AI inference.
• High-speed 400 Gbps Arista networking for low-latency data movement.
• 3 PB Ceph parallel file system with 140 GB/s throughput.

Voyager supports standard AI frameworks like TensorFlow and PyTorch, and is optimized for containerized workloads via Kubernetes.

For more information, visit the Voyager User Guide.

Why Allocations Matter Now

During the 3-year testbed phase, researchers had open access to Voyager. Now, all usage must be tied to an ACCESS allocation.

ACCESS offers several proposal types, including:

• Discover: For small-scale exploratory projects.
• Accelerate: For mid-scale research needing more resources.

Allocations are relatively easy to obtain, especially for AI/ML projects.

Performance Compared to GPUs

Voyager’s Habana Gaudi processors are designed specifically for AI training, offering competitive performance to traditional GPUs. While exact performance comparisons vary by workload, Gaudi accelerators have demonstrated efficiency in large-scale deep learning tasks.

Getting Started

If you’re interested in using Voyager:

1. Submit an ACCESS allocation proposal via https://allocations.access-ci.org/project-types.
2. For assistance, contact SDSC support.

Note: If you have a pre-existing allocation, there is no need to write a new proposal—you can simply transfer existing credits.