Supermicro SYS-822GS-NB3RT AI Server with NVIDIA HGX B300
An enterprise 8U platform built for demanding AI training and inference. Combining the eight-GPU NVIDIA HGX B300 architecture with high-performance CPUs, system memory and data-center networking options.
Product family: SYS-822GS-NB3RT · Gold Series SKU: SYS-822GS-NB3RT-01-G2

Inside the Supermicro SYS-822GS-NB3RT
The 8U system brings together CPU sockets, DDR5 memory, GPU infrastructure, PCIe connectivity and a high-capacity cooling design. The image below illustrates the internal layout and key motherboard components.
Designed for dense compute
The platform layout includes two Intel Xeon CPU sockets, 32 DDR5 DIMM slots, internal fans and PCIe expansion architecture. The HGX B300 baseboard provides high-speed GPU-to-GPU communication through NVIDIA NVLink and NVSwitch.
- 32 DDR5 ECC RDIMM slots
- Dual Intel Xeon 6700-series processor support
- High-capacity system cooling architecture
- 8U rackmount chassis
Supermicro SYS-822GS-NB3RT Specifications
The table separates platform-level capabilities from the components of a specific preconfigured SKU. Final specifications depend on the exact system configuration and bill of materials.
| System family | Supermicro SYS-822GS-NB3RT |
|---|---|
| Configured SKU reference | SYS-822GS-NB3RT-01-G2 Gold Series |
| Chassis | 8U rackmount GPU server, air-cooled platform |
| GPU platform | NVIDIA HGX B300, eight-GPU SXM platform |
| GPU memory | Confirm the exact GPU memory specification against the current NVIDIA HGX B300 platform documentation and final system BOM. |
| GPU interconnect | NVIDIA NVLink and NVSwitch high-speed GPU fabric |
| CPU architecture | Dual-socket Intel Xeon 6700-series support; exact CPU model is SKU-dependent |
| System memory | 32 DDR5 DIMM slots; supported capacity and speed depend on DIMM type and population |
| Storage | NVMe storage options; drive count and capacity depend on the selected SKU |
| Networking | High-speed data-center networking options; adapter and port configuration are SKU-dependent |
| Expansion | PCIe expansion architecture; verify available slots and risers against the final configuration |
| Power and cooling | Redundant power and system cooling designed for high-density GPU operation; confirm exact PSU configuration and input requirements before deployment |
Eight-GPU NVIDIA HGX B300 Architecture
NVIDIA HGX B300 is designed for demanding AI workloads that benefit from tightly coupled GPU compute, high-bandwidth memory and fast communication between accelerators. In this system family, the eight-GPU HGX platform provides the foundation for large-scale AI infrastructure.
Large-Scale AI Training
Build infrastructure for training and fine-tuning large language models, multimodal systems and other compute-intensive deep learning workloads.
Enterprise AI Inference
Deploy generative AI services, production inference pipelines and concurrent model workloads on a high-density GPU platform.
NVLink and NVSwitch are designed to provide high-speed communication between GPUs. Actual application performance depends on model architecture, software stack, data pipeline, precision and workload configuration.
System Memory, NVMe Storage and Data-Center Networking
GPU performance is only one part of an AI infrastructure design. System memory, storage throughput and network topology also affect data loading, distributed training and production inference.
DDR5 System Memory
The 32-DIMM platform provides room for high-capacity DDR5 memory configurations. Supported memory speed and maximum capacity depend on processor selection, DIMM type and memory population rules.
High-Speed Networking
Network adapter options should be selected to match the required bandwidth, fabric protocol, switch compatibility, cable type and cluster topology.
For multi-node deployments, validate the complete network design—including switches, transceivers, cables, routing and software configuration—before ordering the system.
Where Can the HGX B300 Server Be Used?
- Large language model training, fine-tuning and inference
- Generative AI and multimodal model development
- Computer vision, image and video analytics
- Scientific research, simulation and high-performance computing
- Large-scale data analytics and financial workloads
- Medical research, life sciences and computational biology
- Private AI infrastructure for enterprises and data centers
Deployment Considerations Before You Buy
High-density GPU servers require careful planning across rack space, power delivery, cooling, networking and operations. Review these requirements with your data-center team before deployment.
Rack & Power
Check the 8U chassis dimensions, system weight, input power requirements, PDU capacity, circuit design and redundancy plan for the final SKU.
Cooling & Airflow
Validate room conditions, hot-aisle/cold-aisle design, airflow direction and heat-removal capacity for the expected workload.
Network Fabric
Confirm switch compatibility, link speeds, cable or optical module selection, fabric topology and cluster connectivity.
Software & Support
Plan operating system, drivers, AI software stack, system management, warranty and support requirements before procurement.
Gold Series SKU: SYS-822GS-NB3RT-01-G2
The -01-G2 designation refers to a specific preconfigured system SKU within the SYS-822GS-NB3RT family. Component-level details should always be checked against the current manufacturer datasheet and the final quotation.
| System model | SYS-822GS-NB3RT-01-G2 |
|---|---|
| GPU platform | NVIDIA HGX B300 eight-GPU platform |
| CPU | Confirm exact processor model and quantity against the latest -01-G2 datasheet |
| Memory | Confirm installed capacity, DIMM type and speed against the final BOM |
| Boot storage | Confirm installed M.2 / NVMe drive configuration |
| Networking | Confirm adapter model, port count and link speed for the ordered SKU |
| Power | Confirm PSU quantity, rating, redundancy and input requirements |
Supermicro SYS-822GS-NB3RT Price and Availability
The price of a Supermicro HGX B300 server depends on the exact system SKU, processor selection, memory capacity, storage, network adapters, power configuration, support coverage and supply conditions. A project-specific quotation is recommended for an accurate price and lead time.
Sadid Gostaran can help review your workload requirements, identify a suitable configuration and coordinate procurement for enterprise AI and data-center infrastructure.
Frequently Asked Questions
What workloads is the Supermicro SYS-822GS-NB3RT designed for?
It is designed for demanding AI training and inference, large language models, generative AI, data analytics, scientific research and high-performance computing.
How many GPUs are included in the HGX B300 platform?
The HGX B300 platform used in this system family is an eight-GPU architecture. Confirm the exact ordered configuration on the final system quotation.
Does every SYS-822GS-NB3RT have the same components?
No. CPU, memory, storage, networking and power configurations can vary by SKU. Check the full part number and bill of materials before ordering.
Can the system memory be expanded?
The platform provides 32 DDR5 DIMM slots. Supported capacity and speed depend on the processor, DIMM type and population rules.
How can I request pricing and lead time?
Use the product link to contact Sadid Gostaran. The sales team can confirm the required configuration, availability and quotation details.
Manufacturer References
For the latest platform details and configuration-specific information, refer to the manufacturer resources below.
Planning Your HGX B300 Infrastructure?
Contact our team to review your workload, confirm the right system configuration, and request pricing and availability.
View Product & Contact SalesAvailable in Iran
Looking for the Supermicro HGX B300 product page for the Iranian market? Visit the Persian product page for local product information and purchasing options.
View Persian Product PageExplore more: NVIDIA product directory · AI infrastructure in Iran