1569954 : Request for Access and Allocation on DESY GPU Cluster for 2HDM Lattice Project - Francisco Arco

Created: 2026-01-22T10:12:52Z - current status: new

Here is the anonymized and summarized version of the request:

Summary of Request

A postdoctoral researcher at [INSTITUTION] is working on a project involving large-scale lattice field theory simulations to study non-perturbative phenomena in the Two Higgs Doublet Model (2HDM). The project aims to: - Compute observables from lattice simulations to determine the theory’s spectrum. - Analyze scattering amplitudes using the Lüscher formalism. - Requires GPU-enabled computing resources for production runs and scattering studies.

The researcher requests: 1. Access to the GPU cluster at [INSTITUTION]. 2. Allocation of computing time for the simulations.

They are unsure about the allocation process (e.g., whether GPU time is open or project-based) and are prepared to provide additional details, including: - System sizes, performance expectations, and scaling benchmarks. - Further project or code descriptions if required.

Suggested Solution/Next Steps

  1. Access Request:
  2. The user should email [maxwell.service@desy.de] to formally request access to the Maxwell cluster, specifically the compgpu partition (since the hpcgwgpu partition is reserved for industry collaborations).

  3. They should briefly explain the application (lattice field theory simulations for 2HDM) and confirm that the code allows commercial use (required for compgpu/hpcgwgpu partitions).

  4. Resource Allocation:

  5. The compgpu partition has limits (max 2 concurrent jobs/user, max 2 jobs/user total). If larger allocations are needed, the user may need to discuss alternatives with the Maxwell team.

  6. If the project is academic/research-focused, the user should clarify this in their request, as the compgpu partition is shared with industry users (though preemption rates are expected to be low).

  7. Additional Information:

  8. The user should prepare details about:

    • Code compatibility (e.g., open-source tools like PyTorch/TensorFlow, as proprietary software like MATLAB is unavailable).
    • Expected resource usage (e.g., GPU hours, memory requirements).
    • Project timeline (if applicable).

  9. Acknowledgement:

  10. If granted access, the user should acknowledge the Maxwell cluster in publications (see documentation).

Sources Used

  1. Maxwell Cluster Access Documentation
  2. New GPU Hardware and Partitions (2025-09-29)