Gefion HPC (Partner)

The Gefion HPC is a national-level facility that can support both Tier 2 and some Tier 1 workloads. It is particularly suitable for large-scale model training and distributed computing tasks.

Playground Compute Vouchers WIP

P1 has allocated a portion of its compute budget to support researchers through “Gefion Playground” vouchers. These vouchers enable researchers to test their compatibility with the Gefion HPC system and explore the possibilities of large-scale computing.

Important

These vouchers are not intended for full-scale experiments - they are for trial usage, compatibility testing, and exploring large-scale computing capabilities. Once you’ve confirmed your setup works and evaluated the potential, you’ll need to secure your own project funding for production workloads.

1 Submit Proposal 2 Await Review 3 Begin Onboarding

Submit Project Proposal

Forward a copy of your project proposal for a quick assessment.

• Briefly describe your project, data and workload with clear objectives and scope
• Include specific compute requirements and estimated timeline
• Have you applied for or received NNF credits or similar?
• Would you be willing to share a link to your project repository? While completely optional, this helps us provide more personalized guidance and tailored support for your specific use case

Wait for Review

We'll assess your proposal and project requirements.

• Assessment of alignment with P1 research priorities
• If we have access to your project we can gain more detailed insights into how to best support your when onboarding Gefion
• Contacting Gefion if approved to initiate the onboarding process

Start Onboarding

Once approved, you'll receive access credentials and onboarding instructions to begin using Gefion.

• Receive access credentials and login instructions
• Complete technical onboarding with Gefion team
• Set up your development environment
• Begin testing your code with compute voucher credits

Start Applying

Support

Technical Support

For technical issues, contact the DCAI Support Team

Policy Support

For policy issues, contact compute-governance-p1@aicentre.dk

General Questions

Use the #compute or #ask-compute channels on P1 Slack

Compute Coordinator

Contact bstja@dtu.dk for general or technical compute-related questions

Specification

• NVIDIA DGX SuperPOD
• Multiple DGX nodes (8 x H100s) for a total of 1.528 H100 GPUs with high-performance interconnects
• Large-scale GPU resources for distributed training
• High-bandwidth storage solutions
• Scheduling Environment: SLURM