Skip to content
Spike Documentation

Publications and communications

Credit the infrastructure when Spike materially supported a result. Keep the acknowledgement concise, then put the reproducibility details in the methods or computational-resources section.

Section titled “Recommended acknowledgement”

Use this wording unless the project, funder, journal, or SCC service owner has approved a required alternative:

Computational resources for this work were provided by the Spike-1 GPU cluster, operated by the Supercomputing Center at Eindhoven University of Technology.

If an SCC engineer made a substantial intellectual or technical contribution, discuss normal authorship or contributor-credit rules with them rather than expanding the infrastructure acknowledgement automatically.

Describe Spike use in a paper

Section titled “Describe Spike use in a paper”

Record enough detail for another researcher to understand the computational conditions:

  • Spike-1 as the execution platform and Run:ai as the workload orchestrator;
  • GPU model and count, worker count, and the approved compute-resource preset;
  • immutable Harbor image tag or digest;
  • framework, CUDA, driver, and important library versions;
  • model, dataset, and code revisions;
  • precision, distributed strategy, random seeds, and checkpoint policy; and
  • the measurement period and any material limitations.

A methods sentence can follow this pattern:

Experiments were executed on the Spike-1 GPU cluster at Eindhoven University of Technology using [GPU count and model] managed through Run:ai. Workloads used container image [immutable tag or digest] with [framework and version]; the complete configuration and code revision are reported in [repository or supplement].

Citation versus acknowledgement

Section titled “Citation versus acknowledgement”

This documentation does not define a DOI or bibliographic record for Spike. Do not invent one. Acknowledge the platform, describe the computational setup, and separately cite the software, models, datasets, and methods used by the research. Ask SCC before submission if a funder or facility-specific statement is required.

Prepare a SPIKE press release

Section titled “Prepare a SPIKE press release”

Contact the SCC service owner and TU/e Communications before announcing a result that names Spike, SCC, TU/e, MDCS, NVIDIA, or another partner. Agree on the publication date, spokesperson, partner attribution, image rights, and which performance or capacity claims have been verified.

Build the draft from five checked facts:

  1. Research purpose: the problem and why GPU computing mattered.
  2. Result: what was demonstrated, without overstating causality or impact.
  3. Spike contribution: the workload type and the practical capability the platform provided.
  4. People and partners: approved names, roles, funding, and quotations.
  5. Reproducibility and responsibility: where readers can find the paper, code, data statement, or contact, subject to project policy.

Avoid claims such as “unlimited,” “dedicated,” “instant,” or “record-breaking” unless the evidence and comparison have been reviewed. Spike is a shared research platform; temporary priority or a successful benchmark is not a permanent platform guarantee.

Draft structure

Section titled “Draft structure”
  • Headline: outcome first; mention Spike only when it adds meaning.
  • Opening: one paragraph covering who, what, and why it matters.
  • Research detail: the method and evidence in plain language.
  • Infrastructure detail: one paragraph explaining Spike’s contribution.
  • Approved quote: named speaker, role, and organization.
  • Close: paper or project link, funding, acknowledgement, and media contact.

Do not publish project names, user identities, unpublished results, sensitive data, screenshots, cluster identifiers, or utilisation figures without the owner’s explicit approval.