SSN_-_Amendment_1_2.pdf

United States Department of the Interior

INTERIOR BUSINESS CENTER

Washington, DC 20240

Computational Infrastructure — GPU Server Node

Amendment 1 provides answers to questions, as follows:

1. Is the 100TB storage requirement raw or usable capacity, and is there a preferred RAID level?

The requested 100 TB refers to usable storage capacity after RAID configuration. The preferred configuration is enterprise-class RAID 6 (or equivalent fault-tolerant architecture) with hot-spare capability to ensure high availability, data integrity, and protection for high-content imaging, AI model development, multimodal characterization, and autonomous laboratory datasets. The system should be scalable to 250–500 TB as future autonomous laboratory operations expand.

2. Do they have a preference for Intel or AMD, or a minimum CPU core count?

There is no preference between Intel or AMD, provided the platform is enterprise-class and optimized for AI, high-throughput image analysis, and laboratory automation workloads. The preferred configuration is 64–128 physical CPU cores to support concurrent robotic workflow orchestration, laboratory information management (LIMS), AI inference, image processing, database services, virtualization, and multiple simultaneous users. The architecture should also support future CPU expansion.

3. How much system memory (RAM) is required?

The preferred configuration is 1 TB of ECC RAM, with support for future expansion to 2 TB or greater. This capacity will support large-scale AI/ML workloads, high-content image analysis, multimodal data integration, virtualization, digital twin applications, and simultaneous execution of autonomous laboratory workflows.

4. What networking is required (10Gb, 25Gb, or 100Gb)?

The preferred networking configuration is dual redundant 100 Gb Ethernet connectivity, with compatibility for existing NIH infrastructure. The system should support high-bandwidth data transfer between robotics, imaging systems, AI compute resources, institutional storage, high-performance computing environments, and future cloud integration. Multiple 25 Gb and 10 Gb interfaces for instrument connectivity are also desirable.

GPU Requirement: Enterprise-class NVIDIA GPU acceleration (or equivalent), preferably supporting 2–4 high-performance GPUs (e.g., NVIDIA L40S, RTX 6000 Ada, H100, or equivalent) with sufficient PCIe capacity and power to support future expansion. The system should be capable of running multiple concurrent AI inference and training workloads for imaging, computer vision, and autonomous laboratory applications.