Final_SOR_CTCMS_FY25_CPUGPU_ar.pdf

PDF 142 KB Posted

Attached to
HIGHLY PARALLEL HIGH-PERFORMANCE COMPUTER SYSTEMS Federal contract opportunity
Solicitation number
1333ND25QNB640484
Issued by
Department of Commerce National Institute of Standards and Technology

About this file

This Statement of Requirements outlines NIST's acquisition of high-performance computing systems for the Center for Theoretical and Computational Materials Science (CTCMS). The requirement includes four line items: (1) four highly-parallel high-performance computer systems with 96-core CPUs, 8 GB RAM per core, and specific technical requirements; (2) one high-performance general-purpose GPUs system with 64 CPU cores and two Hopper-architecture GPUs; (3) one high-performance low-latency storage system with 24 CPU cores and 15 TB NVMe storage; and (4) an optional line item for up to four additional identical highly-parallel compute nodes.

Delivery is specified for NIST's Gaithersburg, MD campus within 30 days of award, with the Subject Matter Expert (SME) verifying system components and performing boot-up testing for acceptance. The vendor will provide a three-year warranty covering replacement of defective parts, including shipping costs. The systems must operate on Linux kernel 5.10 or later, be air-cooled, use 208V single-phase power, and be rack-mountable. No operating system or software will be provided by the vendor, as NIST will supply those components separately.

View the file

Other files for this federal contract opportunity

Other files attached to HIGHLY PARALLEL HIGH-PERFORMANCE COMPUTER SYSTEMS, newest first.
File Type Posted
Provisions and Clauses.pdf PDF

On GovTribe

Work with this file on GovTribe

  • Download the original file
  • Contacts named in this file
  • Similar government files
  • Ask GovTribe AI about this file

Text version

STATEMENT OF REQUIREMENTS

CTCMS Highly Parallel Computer Systems

I. BACKGROUND INFORMATION

The Materials Science and Engineering Division’s Center for Theoretical and Computational Materials Science (CTCMS) operates a research computing cluster in support of computational materials research. This cluster consists of several highly-parallel computer systems, general-purpose graphical processing unit (GPU) computer systems, and associated support equipment, including data storage systems, power conditioning equipment, racks, switches, and cabling. The division is refreshing a portion of this cluster and has a requirement for additional high-performance computer systems.

The requirement consists of (a) four (4) highly-parallel high-performance computer systems, (b) one high-performance general-purpose GPU computer system bearing two GPU devices, and (c) one high-speed data storage system.

These systems will be inserted into existing cabinets and use existing connection infrastructure, so equipment racks and cabling are not included in this requirement.

II. SPECIFICATIONS

The Contractor shall provide computer systems according to the specifications below.

Line Item 0001: BASE LINE ITEM – Highly-parallel high-performance computer systems Four (4) required. The high-performance computer systems must each provide the following capabilities:

• One Central Processing Unit (CPU) containing no fewer than 96 physical cores, with hardware support for at least 192 simultaneous independent execution threads. The CPU must achieve a base clock rate of 2.4 GHz or higher, peak single-core clock rate of 3.7 GHz or faster, and must implement the amd64/ia64t instruction set.

• Installed memory capacity of at least 8 GB per physical core, consisting of error-correcting registered RAM rated at DDR5-6400 or faster. Since our workloads are limited by memory bandwidth and speed, the memory configuration must be balanced across CPU memory channels.

• Two 960 GB capacity or larger solid-state drives, rated for data throughput of 6 GB/s or faster.

• One HDR-capable InfiniBand host channel adapter with a QSFP-type connector.

• Connectivity for a keyboard and local console display. Keyboard connectivity must be either USB or PS/2, and video connectivity must be VGA or better, able to drive a text console.

• At least two Ethernet ports provided through RJ-45 connectors, rated at 1 Gb/s or faster.

• A motherboard capable of driving the CPU, RAM, and other peripherals at their full rated speed.

• A chassis housing all the components, able to operate on 208V single-phase electrical power.

All components must operate at their full rated speeds. The chassis must be a rack-mount chassis and may not exceed 1U per node in height. For a single-node chassis, this requirement restricts the chassis height to 1U. If multiple compute nodes are housed in a single chassis, then the ratio of total chassis height divided by number of nodes must not exceed 1U. The chassis must include rack-mount rails and hardware for standard 19-inch computer equipment racks.

• The system will be air-cooled and must not require chilled water.

• All hardware must operate correctly at full rated speed in the Linux operating system, under kernel version 5.10 or later.

• The contractor must complete burn-in testing of all components prior to delivery.

• The nodes or multi-node chassis must be delivered fully assembled.

• No operating system or software is required. The government will provide the operating system and application software.

Line Item 0002: BASE LINE ITEM – High-performance GPU computer system One (1) required. The high-performance general-purpose GPU system must provide the following capabilities:

• A Central Processing Unit (CPU) containing no fewer than 64 physical cores, with hardware support for at least 128 simultaneous independent execution threads. System configurations with multiple CPUs (multiple sockets) must total least 64 physical cores per socket. The CPU must achieve a base clock rate of 2.25 GHz or higher, peak single-core clock rate of 3.7 GHz or faster, and must implement the amd64/ia64t instruction set.

• Two (2) graphics processing units (GPUs), each implementing the Hopper GPU architecture, with at least 16,896 CUDA cores, and at least 141 GB of HBM3e GPU memory. The GPUs must be connected to the motherboard by PCIe and be air cooled.

• Installed main memory capacity of at least 8 GB per physical core, consisting of error-correcting registered RAM rated at DDR5-4800 or faster, and equally balanced across the

CPU memory channels.

• Two 960 GB capacity or larger solid-state drives, rated at 6 GB/s or faster.

• One HDR-capable InfiniBand host channel adapter with a QSFP-type connector.

• Connectivity for a keyboard and local console display. Keyboard connectivity must be either USB or PS/2, and video connectivity must be VGA or better, able to drive a text console.

• At least two Ethernet ports provided by RJ-45 connectors, rated at 1 Gb/s or faster.

• A motherboard capable of hosting the CPU, RAM, GPUs, and other peripherals at their full rated speed.

• A chassis housing all the components, able to operate on 208V single-phase electrical power, and no greater than 6U in height. All components must operate at their full rated speeds. The chassis must include rack-mount rails and hardware for standard 19-inch computer equipment racks.

• The system will be air-cooled and must not require chilled water.

• All hardware must operate correctly at full rated speed in the Linux operating system, under kernel version 5.10 or later.

• The contractor must complete burn-in testing of all components prior to delivery.

• The node chassis must be delivered fully assembled.

Line Item 0003: BASE LINE ITEM – High-performance low-latency storage system

One (1) required. The storage system must provide the following capabilities:

• A Central Processing Unit (CPU) containing no fewer than 24 physical cores, with hardware support for at least 48 simultaneous independent execution threads. The CPU must have a base clock rate of 2.25 GHz or higher, peak single-core clock rate of 3.7 GHz or faster, and must implement the amd64/ia64t instruction set.

• Memory totaling at least 256 GB, consisting of error-correcting registered RAM rated at

DDR5-4800 or faster, and equally balanced across the CPU memory channels.

• One SSD hard drive of at least 256 GB raw capacity, for the system OS.

• NVMe-attached hard drive or hard drives totaling 15 TB in raw unformatted capacity. Drives must be hot-swappable.

• One HDR-capable InfiniBand host channel adapter with a QSFP-type connector.

• Connectivity for a keyboard and local console display. Keyboard connectivity must be either USB or PS/2, and video connectivity must be VGA or better, able to drive a text console.

• At least two Ethernet ports provided by RJ-45 connectors, rated at 1 Gb/s or faster.

• A motherboard capable of hosting the CPU, RAM, GPUs, and other peripherals at their full rated speed.

• A chassis housing all the components, able to operate on 208V single-phase electrical power.

All components must operate at their full rated speeds. The chassis must include rack-mount rails and hardware for standard 19-inch computer equipment racks.

• The system will be air-cooled and must not require chilled water.

• All hardware must operate correctly at full rated speed in the Linux operating system, under kernel version 5.10 or later.

• The contractor must complete burn-in testing of all components prior to delivery.

• The node chassis must be delivered fully assembled.

Line Item 0004: SEPARATELY PRICED OPTION LINE ITEM – Additional Highly-parallel high-performance computer systems

The contractor shall provide up to four (4) additional highly-parallel compute nodes configured identically to Line Item 0001.

III. DELIVERY

Delivery shall be made to NIST’s Gaithersburg campus, 100 Bureau Drive, Gaithersburg MD and shall occur within thirty (30) days from award of this requirement.

IV. DELIVERABLES

Description Quantity Recipient Due Date

Highly-parallel high-performance computer systems.

4 SME 30 days after award of this requirement.

High-performance general-purpose GPU computer system.

1 SME 30 days after award of this requirement.

High-performance file storage computer systems.

1 SME 30 days after award of this requirement.

Optional first additional highly-parallel high-performance computer system.

1 SME 30 days after award of this requirement.

Optional second additional highly-parallel high-performance computer system.

1 SME 30 days after award of this requirement.

Optional third additional highly-parallel high-performance computer system.

1 SME 30 days after award of this requirement.

Optional fourth additional highly-parallel high-performance computer system.

1 SME 30 days after award of this requirement.

V. ACCEPTANCE

Upon delivery of the equipment to NIST, the Subject Matter Expert (SME) will verify that the correct type and number of components are present and will operate the systems through boot-up. Once all systems have been booted successfully, the order will be accepted.

VI. WARRANTY

The vendor shall provide a three-year warranty on equipment. The vendor will provide replacements for defective parts, including the cost of shipping replacement parts to the NIST site. Defective parts will be returned to the vendor at NIST's cost. The warranty shall commence upon acceptance of the system by the Government.

File details come from the government source that posted it. Updated .