{"resourceId":"julich-jubis-sparxity-sparse-ml-system-20261028","versions":[{"version":"external-5d4a37988e9c5d5db360e41eecda6c7cefabdc17d749be635322cfa02669c6fa","resource":{"id":"julich-jubis-sparxity-sparse-ml-system-20261028","title":"Jülich adds JUBIS, a 656,640-core Sparxity (SpiNNaker2) system beside its exascale Jupiter Booster for sparse machine-learning workloads","organization":"Jülich Supercomputing Centre (JSC), with Sparxity (formerly SpiNNcloud)","sector":"Research computing and HPC architecture","geography":"Germany; international, limited transferability","publishedAt":"28 October 2026 (HPCwire)","publicationDate":"2026-10-28","eventDate":null,"sourceName":"HPCwire","sourceLabel":"Alex Woodie, \"Jülich Gets Sparse Computing Boost from Sparxity Accelerators,\" HPCwire, 28 Oct 2026","sourceUrl":"https://www.hpcwire.com/2026/10/28/julich-gets-sparse-computing-boost-from-sparxity-accelerators/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["infrastructure","governance-procurement"],"finding":"HPCwire reports that JSC announced it has tapped Sparxity, a Dresden and Manchester neuromorphic spinoff, to provide JUBIS, a multi-million-dollar system of 90 boards (48 SpiNNaker2 chips per board, 152 cores per chip) totalling 656,640 cores. JUBIS will sit near the Jupiter Booster, Europe's first exascale system, and run sparse machine-learning workloads such as sparse solvers and optimizers that are a poor fit for Jupiter or other JSC systems. A JSC neuroscience lab scientist said it complements JSC infrastructure and opens new research opportunities. Sparxity's CEO argued that GPUs handle sparse data poorly and that increasingly sparse mixture-of-experts models suit its architecture, and said it has dozens of systems at 30 customers including Sandia National Laboratories.","sledRelevance":"Shows a leading research computing center adding a non-GPU accelerator for workloads GPUs serve poorly, relevant to U.S. centers deciding whether to diversify beyond GPU clusters for sparse scientific ML and MoE inference.","evidence":"Read the full article. Hardware counts and customer claims come from Sparxity via HPCwire; no benchmarks, energy figures or JSC workload results were reported.","architectureImplications":"Heterogeneous centers may place specialized sparse or event-based accelerators beside GPU systems and route workloads by data structure.","governanceImplications":"Adding novel architectures needs workload-matching policies and evaluation before allocation to users.","securityPrivacyImplications":"Not discussed.","caveats":"Vendor-sourced figures relayed by trade press; no performance or efficiency data. Exact cost not disclosed. Efficiency and MoE claims are the CEO's. European context limits direct transfer.","streamIds":["research"],"roles":{"sales":"Lead with workload fit, not hardware novelty. HPCwire reported on 28 October that Jülich is adding JUBIS, a 656,640-core Sparxity system, beside its exascale GPU booster for sparse ML workloads GPUs handle poorly. U.S. research computing directors with sparse solver or MoE inference demand may want an architecture review. Offer a bounded workload assessment. Avoid efficiency claims; the article gives no benchmarks and the figures come from the vendor.","engineering":"Frame fit around routing sparse workloads to suitable hardware. JUBIS is event-based, with many low-power interconnected cores, and is meant for sparse solvers, optimizers and sparse models. Evaluating similar hardware needs representative sparse workloads, a software porting path and scheduler integration. Validate by benchmarking time-to-solution and energy against existing GPU and CPU nodes, since no performance data was published.","delivery":"Treat delivery as a controlled evaluation of a specialized accelerator. Work covers workload selection, porting, benchmarking, user support and scheduler integration. Ownership sits with the research computing center. Governance checkpoints include vendor viability review and go/no-go criteria before allocations. Acceptance criteria are proposals, such as measured energy per solution versus GPU baselines, not results Jülich has reported."},"roleDetails":{"sales":{"customerProblem":"Some scientific ML and inference workloads run inefficiently on GPU clusters, but centers lack evidence on alternatives.","stakeholders":["Research computing director","HPC architects","Faculty with sparse solver or MoE workloads","Procurement"],"discoveryQuestions":["What share of workloads are sparse or irregular?","Are GPU utilization rates low for those jobs?","Is there budget for an evaluation system?"],"valueHypothesis":"A workload assessment may identify jobs where specialized hardware could improve efficiency, to be proven by benchmarks.","potentialEngagement":"A workload characterization and benchmark plan for sparse jobs across current and candidate architectures.","unsupportedClaims":["Efficiency or speed gains","Suitability for general AI training","Vendor customer counts as validation"]},"engineering":{"fit":"Fits centers with sparse solvers, optimization or MoE inference workloads alongside GPU systems.","architectureIntegration":"Specialized event-based accelerator system placed beside GPU partitions, with scheduler routing by workload type.","prerequisites":["Representative sparse workloads","Porting tools and libraries","Scheduler integration"],"constraints":["No published benchmarks","Small vendor ecosystem","New programming model"],"security":"Apply standard cluster access controls and review vendor software supply chain.","proofOfValue":["Benchmark time-to-solution on sparse solvers","Measure energy per solution against GPU nodes","Assess porting effort per application"]},"delivery":{"work":"Select workloads, port and benchmark, integrate with the scheduler, and train users.","dependencies":["Vendor support","Application developer time","Evaluation hardware access"],"ownership":"Research computing center.","skillsAdoption":"Staff and users learn event-based programming models and workload triage.","governanceCheckpoints":["Vendor viability review","Benchmark go/no-go","Allocation policy for specialized hardware"],"acceptanceCriteria":["Energy per solution below GPU baseline on chosen workloads (proposal)","Porting effort documented","User support plan in place"],"risks":["Vendor lock-in or failure","Low adoption due to porting effort","Gains limited to narrow workloads"]}},"retrievedAt":"2026-10-28T22:08:50-05:00","enrichedAt":"2026-10-28T22:08:51-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"New programming models for event-based hardware require user training and support staff skills.","procurementImplications":"Centers can buy specialized accelerators from small vendors for niche workloads, with vendor viability and software support as diligence items.","operatingModelImplications":"Workload triage across architectures becomes a research computing service.","sourceVerification":{"openedUrl":"https://www.hpcwire.com/2026/10/28/julich-gets-sparse-computing-boost-from-sparxity-accelerators/","referenceExcerpt":"The JUBIS system being deployed at JSC contains 90 boards across several racks, representing 656,640 SpiNNaker2 cores in the multi-million-dollar deployment.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}