Huang Complex Supercomputing Seed Fund

Timeline

• Applications Due: June 18, 2025
• Award Period: October 1, 2025 – October 30, 2027

Priorities for Investment

Because this program seeks to help OSU researchers advance PWS goals, competitive proposals should demonstrate alignment with one of the PWS focus areas:

• Climate Science and related solutions
• Clean Energy and related solutions
• Robotics and AI
• Integrated Health and Biotechnology
• Or, building foundational research strength in artificial intelligence, data science, research computing, and/or the social sciences, arts and humanities.

Technical Design of the Huang Complex Supercomputer

The supercomputer will be a NVIDIA SuperPOD with Vera Rubin GPUs. The SuperPOD may be configured as a collection of DGX servers (each with 8 GPUs) and/or as NVL144 systems (multiple racks, each of which has 144 GPUs and 75TB of fast memory). The supercomputer will not be a traditional academic supercomputer that consists of x86 Intel or AMD CPU-based servers with GPU accelerators for enhanced parallel computing. Applications may require modification in order to make full use of the supercomputer. As NVIDIA is increasing its use of ARM processors as CPUs with their GPUs, we expect that the SuperPOD implemented in 2026 will use ARM processors.

NVIDIA SuperPODs use NVLink as their interconnect. NVLink is a 1.8TB/s bidirectional, direct GPU-to-GPU interconnect that scales multi-GPU input and output (IO) within a server. The NVIDIA NVLink Switch chips connect multiple NVLinks to provide all-to-all GPU communication at full NVLink speed within a single rack and between racks. With NV-Switch, NVLink connections can be extended across nodes to create a seamless, high-bandwidth, multi-node GPU cluster—effectively forming a data-center-sized GPU.

Eligibility Requirements

Technical Requirements

The team submitting a proposal must already be running the software application(s) to be scaled up on multiple NVIDIA GPUs. The software application(s) must be compatible with or demonstrated to be effectively transferable to the expected technical design of the Huang Complex supercomputer during the life of the project. They will need to port their software applications to such an environment at the same time as scaling up their applications. During the project, applications must be able to demonstrate scaling up to the use of 8 GPUs prior to qualifying for a hero run.

Other Eligibility Requirements

Teams should have an identified project lead who is an OSU researcher from any OSU campus or facility, with a record of extramurally funded research, previous collaboration, and academic success. Each eligible researcher may lead only one proposal. Researchers may, however, collaborate on multiple proposals. Teams may also include experts external to OSU and apply as collaborative cross-institution partnerships. Projects must already have access to the data needed for testing during the scale-up of their applications.

Award

Maximum budget request per application is $50,000. Successful applicants will be able to request an additional $25,000 after the first year of their project and a demonstration of their team’s success in scaling up their application for a hero run.

Expected Deliverables 

All eligible researchers submitting proposals should note the following parameters and expectations of supported projects:

1. All supported projects will have the name of the principal investigator (PI), their affiliation, project title, and project abstract posted on OSU websites.
2. All project PIs will be added to a mailing list to foster a community of demonstration project participants and their support.
3. Project PIs are expected to provide brief (2–3 paragraphs) project updates at one month and then every three months, with a final report (three pages) upon project completion.
4. For reporting purposes, information will be collected about how all participants use their demonstration project-allocated resources. DRI may share this usage information, typically in aggregate form.
5. Teams are expected to participate in publicity and outreach events about their supercomputer hero runs and their solution’s impact.
6. Because of the strong public component of the Huang Complex, all project results must be open and publishable. Preliminary results must be available for university publicity within 90 days after a hero run.

Application Process

The OSU internal submission platform, InfoReady, will be used for submitting applications. Applicants who do not already have an InfoReady account will need to create one using their ONID. Applicants must complete an online application that includes the following:

• Name, title, and unit of project lead (PI)
• Name, title, and affiliation of other known team members including external team members
• Project title
• Identification of the PWS area the project is most aligned with (see Priorities for Investment above)
• Name of the application to be scaled up
• Brief abstract of the intended project (up to 100 words)
• Total budget requested (up to $50,000)

Along with the application form, the following supplemental documents must be uploaded as part of the completed application package on InfoReady:

1. Required: Narrative Plan (see below for details)
2. Required: Itemized Budget, using required template (see below for details). Pre-award support personnel in your college are essential partners in developing the budget and justification.
3. Required: Detailed Budget Justification, using required template
4. Required: Updated CV/Resume for each committed team member that captures relevant expertise (a two-page CV is recommended)
5. Encouraged, not required: A public repository with test code on GitHub that demonstrates how the team will be using multiple GPUs in parallel (whether using the techniques documented by Nvidia, such as NVLink or NVSwitch, or other methods)

Narrative Plan

Narrative Plan Outline

The proposal narrative should be five to six pages long, using a font no smaller than 11 point and with 1” margins. To ensure your request can be properly reviewed, your Narrative Plan should include the sections outlined here:

Scientific Goals

Your proposal should devote up to two pages to describing the scientific goals, which should be sufficiently detailed to help reviewers understand the rationale for the resources being requested. If your research is not already supported by a merit-reviewed funding grant, you may consider devoting additional space to describing the scientific merit of the work. In particular, describe how your proposed work contributes to the supercomputing seed fund’s Priorities for Investment:

• What are the big scientific challenges being addressed and what big solutions will be sought requiring the use of a hero run on this big computer?
• How would this seed grant funding for your demonstration project support, or relate to, current merit-reviewed grant funding?
• Where do you plan to publish or otherwise publicly disseminate your results, in what digital form, and on what timeline?
• How will the ability to implement your application on the supercomputer in a hero run enable your submission of future proposals for external funding?

Technical Goals and Methods

Your proposal should devote up to three pages to Technical Goals and Methods. The following questions should be addressed in the Technical Plan section of your proposal:

Current Application Status

• What application is the target of your scaling? What is the origin of that application: open-source project, community code, your lab, etc.?
• What is the scale of your current use of your application, including the amount of data, the number and type of GPUs, any associated CPUs, and the length of a run?
• What is the type of server architecture on which the application is being used, e.g., NVIDIA DGX Server, CPU servers with GPU accelerators, or other (describe)?
• What was your experience with scaling up use of your application to the present number of GPUs?

Hero-Run

• What is your goal for the amount of computing time needed from the Huang Complex supercomputer in terms of the number of GPU-hours for a hero run with your application?
• How have you arrived at your estimated resource needs? How does this break down into the number of independent computations and their individual computing requirements? A table is often useful for presenting this information.
• How have you determined the computational costs, whether by prior efforts by the team, citable results from other researchers, or benchmark runs?
• How distributed can the computation be, and can it be split across multiple DGX resources?
• How much storage would be needed for a hero run?

Technical Approach to Scaling

• What are the changes required to your application to scale up to a hero run on the Huang Complex supercomputer?
• What are the technical goals or software engineering obstacles to overcome?
• What is your proposed approach for how you will make and test those modifications?
• How will your scaled-up application run on the SuperPOD architecture, across multiple GPUs and multiple nodes?
• What computing resources will you use, or will you require, during the scaling-up of your application?
• Will your project require access to specific software packages, other than your application? If so, describe this software (including whether it is open source or licensed) and how you intend to use it.
• Will there be any restrictions that might apply to the project, such as export-controlled code, ITAR restrictions, proprietary data sets, or personal health information (PHI) or HIPAA restrictions? In such cases, please provide information on security, privacy, and access issues.  

Project Implementation, Support and GPU Needs

Your proposal should devote up to two pages to its Project Implementation, Support, and GPU Needs section.

• What are the plans and the timetable (including project status and milestones at 1 month and then every three months until the application is ready to run on the Huang Complex supercomputer) for getting to the goals of your project? Describe how you plan to accomplish the proposed work by the opening of the Huang Complex in the Fall of 2026.
• Summarize your team's qualifications and readiness to execute the project in using both the methods proposed and the resources requested.
• For other types of support or assistance, describe whether additional technical assistance from staff, e.g., at NVIDIA, will be essential, helpful, or unnecessary. The estimates of the necessary support are very helpful.
• GPU resources are available through on-campus systems, NAIRR, and commercial cloud services. How do you propose to obtain access to the GPUs needed during scaling up? Do you need assistance with this?

Budget and Budget Justification

Budgets should reflect realistic costs that might be incurred by implementing the scope of work described in the narrative plan. Maximum budget request is $50,000. (An additional $25,000 can be requested later when the annual report is submitted.) No match is required.

Pre-submission consultation with your home college(s) and unit is required. Please reach out to pre-award support in your college as soon as possible for budget development and completion of the budget and budget justification templates. If you do not know who in your college can provide this support, reach out to your Associate Dean for Research. The completed budget and justification must be uploaded to the InfoReady platform for your application to be reviewed. You can download the budget template [insert when available] and the budget justification template [insert link when available] by logging in with your ONID.

Allowable Costs: Allowable costs include but are not necessarily limited to: PI salary and OPE, including summer effort and academic year effort¹ if newly allocated for the purpose of this award; honoraria to external collaborators or advisors; incentives and other costs that support external partner engagement; research support staff including students and postdoctoral scholars who are directly contributing to the scope of work; materials, subscriptions, equipment use associated with producing preliminary results; travel; and meeting facilitation and hosting. Costs must represent new investments or allocations of effort to research development and cannot offset ongoing college or unit costs.

¹Requests for academic year salary do not guarantee course buyouts. Course buyouts must be arranged at the discretion of the department head separate from this application process. We encourage you to talk with your unit leadership about the availability and cost of a buyout before applying.

Not Allowable Costs: Costs for capital equipment, subawards, expenses related to obtaining data needed to test scaling up your application or during a hero run, and indirects are not allowed in the project budget. The Budget Justification must clearly state how each cost will advance efforts to scale up your application, generate viable products and early results, and produce submission-ready proposals based on your computational work.

Review and Approval Process

The proposal package includes a completed application form on InfoReady along with the required supplemental documents as listed above. Proposals will be reviewed by ORA (Office of Research Advancement) to ensure completeness and adherence to these guidelines.

Proposals that are complete and compliant with the guidelines will be reviewed by two panels, one for a scientific review and a second that will perform a technical review. Each panel will have at least two members. Reviewers will be selected to minimize conflicts of interest and optimize capacity to provide actionable feedback to applicants. Reviewers will complete a rubric providing both quantitative scores and qualitative feedback in the following areas:

• Alignment of scientific goals with the stated priorities for investment (see Section II, Priorities for Investment, above)
• Rationale and potential scientific and societal impacts of a hero run
• Need for the computing, support, or other resources available with support from the supercomputing seed fund, to achieve scientific goals
• Feasibility of the technical approach to modifying the team’s application to scale up to a hero run
• Feasibility of the team’s approach to computing that spans multi-GPU and multi-nodes with the scaled-up application
• Applicant has assembled, or provides a feasible plan to assemble, a team able to succeed with the proposed work
• Planned expenditures are well-justified
• Project readiness and potential for near-term progress  
• Feasibility of a hero run with the scaled-up application and its data/storage requirements
• Estimated computational resource requirements and justification

After completion of the independent review process, selected proposers will be notified by ORA of their award and assisted to identify the most appropriate computational resources for the first stage of scaling up their application.

Award set up: Upon selection, awardees will coordinate with their college pre-award staff to submit the budget and budget justification through Cayuse to set up an internal spending index.

Important Dates

• May 29, 2025: 10:00 a.m. Zoom – Online information session
• June 3, 2025: 1:00 p.m. Zoom – Online Information session
• June 18, 2025 – Full applications due on InfoReady
• June 30, 2025 – Panel reviews complete
• July 8, 2025 – Awards notifications
• October 1, 2025 – One year $50,000 award period begins
• November 3, 2025 – One month report due
• October 1, 2026 –Annual report due, with request for second $25,000 supplement
• October 30, 2027 – Final report due, award ends

Important Contacts

For technical questions on the Huang Complex supercomputer or access to GPU resources, contact Dirk Petersen, Director of the Supercomputer Center at [email protected].

General questions from college leadership or unit directors, or from researchers, should be directed to David Barber, Director of Strategy for Research Computing, Office for Research Advancement at [email protected].

For questions interpreting budget guidelines prior to submission, please contact your college’s pre-award support staff. For technical difficulties with the InfoReady system, please click the Help link at the top right corner and submit a support ticket.