From the Research Universities edition of September 25, 2026
RIT TIGRIS: campus HPC cuts average job wait from ~6 hours on SPORC to ~18 minutes; TOP500 #302 / Green500 #82
Rochester Institute of Technology (RIT) Research Computing · Higher education research computing · United States (Rochester, NY)
- Publisher
- RIT News
- Original publication
- September 25, 2026
- Source retrieved
- 2026-09-26
- Event date
- 2026-09-25
What happened
RIT reports that TIGRIS, launched summer 2026, delivers about 3× CPU and 7× GPU performance versus prior cluster SPORC, ranks No. 302 on TOP500 and No. 82 on Green500 (No. 18 / No. 15 among U.S. academic systems on those lists), and that average job wait fell from about six hours on SPORC to about 18 minutes on TIGRIS. Research Computing cites support for 830+ scholarly works since 2016 and shared services across faculty and 1,600+ students.
Why it matters
Campus research computing leaders evaluating AI/HPC capacity upgrades need measured queue-time and ranking outcomes, not only GPU counts, when justifying shared clusters for AI-assisted discovery workloads.
Evidence and measured results
RIT News article dated September 25, 2026 opened via HTTPS; queue-wait, TOP500/Green500, and 3×/7× performance claims taken from page body.
Limitations and uncertainty
Self-reported institutional metrics; no independent replication of queue statistics; ranking snapshot dates not fully specified beyond list membership.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-26; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Campus researchers lose days in HPC queues while AI and simulation demand outstrips legacy shared clusters.
- Stakeholders
- VPR/research administration
- Research computing director
- College deans
- Faculty power users
- CIO/facilities for power/cooling
- Discovery questions
- What is current average/percentile queue wait by partition?
- Which AI vs simulation workloads drive GPU demand?
- Is Research Computing funded as a shared service or recharge-only?
- What TOP500/Green500 or efficiency goals matter to leadership?
- Value hypothesis
- A shared next-gen GPU cluster plus Research Computing services can cut queue waits by an order of magnitude versus a saturated legacy cluster, if utilization and training keep the system shared.
- Potential engagement
- Queue-telemetry baseline and 90-day migration pilot for 3–5 AI/HPC groups with wait-time and utilization SLOs.
- Unsupported claims
- Guaranteed discovery breakthroughs
- Zero cooling/power constraint
- Replacement of all PI-owned GPUs
- Autonomous lab outcomes
Pre-sales engineering
Role takeaway
- Fit
- University-wide ARM/GH200 HPC cluster coexisting with legacy x86 GPU partitions under Research Computing ops.
- Architecture and integration
- Integrate Slurm fair-share, parallel filesystems, AI software modules, identity, and queue telemetry dashboards; keep SPORC-class systems for transitional workloads.
- Prerequisites
- Power/cooling headroom
- Research storage capacity
- Identity and allocation policy
- User training program
- Baseline queue metrics
- Constraints
- Institutional self-report of waits
- Single-campus deployment
- Ranking lists are snapshots
- Does not specify agent safety controls
- Security
- Define data classification, multi-tenant isolation, and export-control paths before hosting sensitive or regulated research on the shared AI partition.
- Proof of value
- Average/percentile job start wait vs prior cluster
- GPU/CPU utilization
- User-reported time-to-result on named workloads
- Published efficiency ranking if claimed
Delivery
Role takeaway
- Work
- Commission cluster, migrate pilot workloads, publish wait-time SLOs, and stand up Research Computing consult/training.
- Dependencies
- Facilities power/cooling
- Network/storage
- Identity/IAM
- PI pilot groups
- Communications for cutover
- Ownership
- Research Computing owns platform ops and queue policy; PIs own scientific methods; VPR owns allocation governance.
- Skills and adoption
- Train users on fair-share, right-sizing GPUs, and module stacks; train consultants to read queue telemetry.
- Governance checkpoints
- Allocation policy sign-off
- Security/data-class review
- Pilot SLO review at 30/90 days
- Acceptance criteria
- Documented wait-time improvement vs baseline
- Service catalog live
- Pilot groups productive on TIGRIS-class partition
- Risks
- Queue metrics regress under load
- PI shadow IT persists
- Power/cooling limits throttle GPUs
- Overclaiming rankings as scientific quality
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Shared GH200-class campus cluster as university-wide batch resource alongside prior SPORC; Research Computing model bundles HPC, storage, training, workflow optimization, and grant facilitation rather than PI-owned islands.
Governance
Who approves, reviews and stays accountable for outcomes?
Institutional ownership of fair-share scheduling and shared capacity implies research-admin policies for priority, training, and allocation—not only capital purchase of GPUs.
Security and privacy
What data, permissions and controls need testing?
University news does not detail tenant isolation, research-data classification, or export controls; treat as capacity announcement requiring local security review before AI-agent or sensitive-data use.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Shared cluster plus Research Computing training expands access for students and faculty who cannot buy private GPUs; still depends on fair-share literacy.
Procurement
What should contracts, pricing and exit terms secure?
Justify shared cluster CapEx/OpEx with wait-time SLOs and service wrapper, not GPU brochure FLOPS alone; include Green500-style efficiency in RFP scoring if energy goals matter.
Operating model
Which teams own the service once it runs?
Research Computing as university-wide service (HPC, storage, consult, grant help) is the operating model that converts hardware into reduced wait times.
Publication history
- 2026-09-25Research Universities · Issue 194 resources
Stable resource ID: rit-tigris-research-computing-20260925