Skip to content
Spike Documentation

Spike documentation

Spike is TU/e’s shared environment for research that needs serious GPU computing. It brings hardware, scheduling, project storage, and reusable container environments together so research teams can spend less time assembling infrastructure and more time running experiments.

Choose your platform path

Section titled “Choose your platform path”

Select your team profile to understand the architecture, storage, and onboarding route for your workloads before exploring the documentation:

Academic Research Track

Spike 1 for TU/e Researchers & Students

For TU/e researchers conducting high-performance AI research. Containing state-of-the-art GPU resources for training, high-performing storage & interconnects between nodes and hosted frontier Large Language Models.

●Compute:Academic NVIDIA B200 Partitions
●Storage:250 TiB High-Performing Project Space (DDN) with S3 Option
●Services:Dedicated AI engineering support, NVIDIA AI Enterprise & hosted frontier LLMs
Your Documentation Journey

Step-by-Step Path from Access to Running Jobs

01
Intake & ConsultationSubmit TOPdesk request & hold 1-on-1 discussion with an SCC AI engineer.
02
Accounts & AccessSign in with TU/e SSO on Run:AI, connect EduVPN, and verify Harbor access.
03
Storage Architecture250 TiB high-performing DDN project space at `/data` with S3 connection options.
04
Data TransferMove datasets efficiently via direct workload downloads, S3 sync, or rsync.
05
Blackwell B200 ComputeDeploy workloads across 8x B200 GPUs (192GB HBM3e) with NVLink mesh & FP4 precision.
06
Intimate AI Engineer SupportDedicated hands-on guidance from SCC AI engineers throughout your research lifecycle.
⚡Academic research workloads operate on dedicated DGX B200 nodes with fair-share GPU scheduling and 250 TiB DDN storage.
Blackwell B200 Specs Start Researcher Onboarding

Common task guides

Section titled “Common task guides”
Getting startedGet accessCheck what you need before submitting work to Spike.Open guide WorkloadsRun a training jobSubmit a reproducible run and follow it from the queue to completion.Open guide ContainersPrepare an environmentPackage your code and dependencies with Podman and Harbor.Open guide Interactive workConnect to a workloadOpen a safe terminal connection to a running workspace.Open guide LLM hosting · Spike 1Use a hosted modelAccess the current SCC-managed LLM gateway for academic research without operating a deployment.Open guide

Built together at TU/e

Section titled “Built together at TU/e”

Spike is a shared university investment, made possible by EAISI and six TU/e departments: Mathematics and Computer Science, Electrical Engineering, Mechanical Engineering, Biomedical Engineering, Industrial Design, and Industrial Engineering & Innovation Sciences.

For commercial and industry computing on Spike A, dedicated NVIDIA Blackwell B300 systems are provided through Supercomputing Brabant and hosted and operated by TU/e as a founding major partner.

That mix matters. Spike supports work across disciplines, from medical imaging and scientific machine learning to design and engineering research. The TU/e Supercomputing Center operates the shared platform and helps turn that collective investment into computing researchers can use.

What Spike provides

Section titled “What Spike provides”

Spike provides shared GPU systems and scheduling, project-scoped data sources, and container-based environments. Access is organized through projects: the project you work in determines which data, environments, and compute resources you can use.

This site covers the researcher-facing path, from getting access to running and troubleshooting workloads. Credentials, internal identifiers, quota decisions, and project-administration procedures stay out of the public documentation.