Tillicum: GPU-Accelerated Research Computing Platform


On This Page

 

Tillicum poster with Top500 and Green500 results.

Data for image pulled on March 19, 2026. Tillicum debuted on the Top500 list in November 2025 at number 185, which is number 7 across supercomputers in United States higher education, and number 51 on the Green500. 
 

Overview

Tillicum is the University of Washington’s next-generation, GPU-accelerated computing platform designed to support cutting-edge research, teaching, and learning in AI, machine learning, data science, and scientific simulation. Operated by UWIT Research Computing, Tillicum provides scalable, high-throughput, GPU-accelerated computing without requiring hardware ownership. 

Tillicum is part of the integrated research cyberinfrastructure at UW, offering a flexible, usage-based access model for time-bound projects requiring GPU performance at scale. 

Named after the Chinook Jargon word Tillicum, meaning "the people", the platform celebrates the collaborative spirit of research at UW and honors the Coast Salish peoples on whose land we learn and work. 

Key Features 

Tillicum is built for GPU-accelerated performance and large-scale scientific workflows. Key features include:

Business Model and Pricing

Tillicum operates under a usage-based allocation model, offering researchers access to GPU-accelerated computing and storage for defined project periods. Researchers only pay for the compute time they used under a pay-as-you-go model. Unlike Hyak Klone, there are no slot fees or hardware purchases - pricing reflects on-demand, project-based compute access. Tillicum is exempt from indirect (F&A) costs. 

Pay-as-you-go by the GPU Hour

On Tillicum, computing units are "GPU Hours" and are applied to all jobs scheduled with Slurm including debugging and testing.

Tillicum Usage Rates
 
GPU Hour = Elapsed Time x N GPUs
 
Usage Rate: $0.90/GPU Hour - Billing is monthly and handled as a subscription in UWIT's ITBill system.
 
Every scheduled job on Tillicum is subject to a the usage rate and requires at least 1 GPU (141GB RAM).
  • Jobs are bound by a maximum of ~200GB system RAM and 8 CPUs.
  • If more system RAM or more CPUs are required, additional GPUs must be added.
 

Tillicum processes not subject to charge:

Budget Usage Limits

Budget usage limits are available to help research groups limit spending on Tillicum. Limits can be added at the account level or at the individual user level.If you chose to provide a usage budget, we have two options for usage budget enforcement: 

Option 1: Warn Only – Jobs that would exceed the limit are still allowed to run, but a warning will appear upon submission. This maintains full workflow continuity while increasing visibility into usage. 

Option 2: Enforced – Jobs that would exceed the defined limit will not run. This ensures your group stays strictly within its established budget.

Storage Policy 

Tillicum provides high-speed, temporary storage designed for active computation. Every user has a home directory by default, project/lab dedicated storage, and scrubbed storage for temporary overflow use.

We will constantly evaluate this policy based on user feedback.

Get this Service

  1. Submit a Tillicum Access Intake Form.
  2. Provide a description of:
    • Research goals
    • Required compute resources
    • Expected timeline

Our team will review your request and get back to you with next steps. Tillicum access is typically delivered within 7 business days after a billing worktag has been received. 

Access to Tillicum

Option A: Free Community Access

All faculty researchers* are eligible for 100 promotional GPU hours to evaluate Tillicum at no cost; no payment information required. This benefit is limited to one project per principal investigator and may only be used once.

Demonstration accounts require the explicit written consent of the faculty researcher. They are subject to a strict 100 GPU hour limit and a maximum of 100 GB of storage. A demonstration account can be converted to a regular account at any time by adding a Workday budget worktag before the promotional hours are exhausted.

Option B: New Regular Accounts

All faculty researchers are eligible for 100 promotional GPU hours to evaluate Tillicum. Enrolling in a regular account requires a Workday budget worktag and includes a 1 TB dedicated storage allocation along with the 100 promotional GPU hours. Once the promotional hours are exhausted, usage will be billed to the provided worktag on a monthly basis.

Option C: Additional Tillicum Accounts

Groups that manage separate projects or funding sources may benefit from maintaining multiple Tillicum accounts. This option is available upon request. Please provide as much detail as possible in the comments section of the enrollment form.

Note that additional storage allocations are not available for additional accounts. Storage is shared across all projects associated with a single faculty researcher.

Intake Form

*"Faculty researcher" is used here as the typical case, but this option is open to any principal investigator of UW-sponsored research. Tillicum accounts may be provisioned to UW affiliates who have a valid UW budget worktag.

Tillicum GPU Credits for Students 

UW students can apply for $250 in Tillicum GPU computing credits to support research projects, class work, and technical exploration that benefit from high-performance GPUs. Tillicum features NVIDIA H200 GPUs and is designed for workloads that can take advantage of large GPU memory or multi-GPU scaling. 

These credits are provided by the Research Computing Club, with funding from Student Technology Fees, to help expand student access to advanced research computing resources. 

Learn more and apply.

Eligibility

A UW worktag is required to enroll in this service. 

Tillicum is available to:

Restrictions

Not supported:

Important

Tillicum is not approved for use with HIPAA-regulated Protected Health Information (PHI), as data stored on Tillicum is not encrypted at rest.

Support and Availability

User support

Service Team Support

Related Services

Questions