100% Original Wholesale brand new SYS822GA NGR3 Dual Intel® Xeon® 6900 series processors with P-cores AI DATA SERIES

Feature	SYS-822GA-NGR3 (2U)	SYS-A22GA-NBRT (2U)
Form Factor	2U Rackmount	2U Rackmount
CPU Support	Dual 4th Gen Intel® Xeon® SP (SPR)	Dual 4th Gen Intel® Xeon® SP (SPR)
Max CPU TDP	350W	350W
Memory	32x DDR5 (8TB max)	32x DDR5 (8TB max)
GPU Support	8x NVIDIA HGX H100 (SXM5)	4x GPUs + 2x BlueField-3 DPUs
GPU Type	NVIDIA SXM5 Form Factor	PCIe Form Factor
Storage	10x 2.5″ NVMe/SATA/SAS	12x 2.5″ NVMe/SATA/SAS
PCIe Expansion	8x PCIe 5.0 (GPU-optimized)	6x PCIe 5.0 (4x GPU + 2x DPU)
Networking	4x 200G NVIDIA ConnectX-7 (Standard)	2x 200G ConnectX-7 + BlueField-3 DPU
Acceleration	Pure GPU Compute	GPU+DPU Hybrid Acceleration
Power Supply	3200W Titanium (N+N Redundant)	2200W Titanium (N+1 Redundant)
Cooling	Direct Liquid Cooling (DLC) Ready	Air/Liquid Hybrid Cooling
Ideal Workloads	AI Supercomputing, Large Model Training	Secure AI, Cloud-Nati

Supermicro SYS-822GA-NGR3 – 2U 8-GPU AI Supercomputing Server

Ultimate AI Performance in a 2U Form Factor

The Supermicro SYS-822GA-NGR3 represents the pinnacle of AI-optimized server technology, delivering unprecedented GPU density in a standard 2U chassis. Designed specifically for large-scale AI training and HPC workloads, this powerhouse system supports 8x NVIDIA HGX H100 SXM5 GPUs with NVLink connectivity, enabling breakthrough performance for the most demanding artificial intelligence and deep learning applications.

Key Features:

1. Unmatched GPU Acceleration

• 8x NVIDIA H100 SXM5 GPUs with NVLink (700W each)

• Full GPU-to-GPU interconnect via NVLink Switch System

• Supports NVIDIA HGX H100 8-GPU Baseboard

• Optimized for large language models (LLMs) and generative AI

2. Dual Intel Xeon Scalable Processors

• Dual 4th Gen Intel® Xeon® Scalable processors (Sapphire Rapids-SP)

• 32x DDR5 DIMM slots (8TB max memory)

• PCIe 5.0 architecture throughout

3. High-Density Storage Options

• 10x 2.5″ hot-swap drive bays (all NVMe capable)

• Support for EDSFF E3.S storage devices

• Hardware RAID options available

4. Advanced Networking

• 4x 200Gb/s NVIDIA ConnectX-7 SmartNICs (standard)

• OCP 3.0 mezzanine slot for additional networking

• 1GbE management port

5. Extreme Power & Cooling

• 3200W Titanium-level redundant power supplies

• Direct Liquid Cooling (DLC) ready

• Optimized airflow for sustained GPU workloads

• N+1 fan redundancy

6. Enterprise-Grade Management

• Supermicro BMC with Redfish API

• NVME-MI support for storage monitoring

• GPU telemetry and monitoring

Ideal Workloads:

• Large-scale AI model training (LLMs, generative AI)

• HPC and scientific computing

• Deep learning inference at scale

• 3D rendering and simulation

• Financial modeling and quantitative analysis

Why Choose the SYS-822GA-NGR3?
✔ Industry-leading 8-GPU density in 2U
✔ Full NVLink connectivity for maximum GPU-to-GPU bandwidth
✔ Liquid cooling ready for datacenter efficiency
✔ Optimized for NVIDIA AI Enterprise software stack
✔ Supermicro’s proven rack integration expertise

Technical Specifications Summary:

• Form Factor: 2U rackmount

• CPU: Dual 4th Gen Intel Xeon Scalable

• GPUs: 8x H100 SXM5 (700W each)

• Memory: 32x DDR5 DIMMs

• Storage: 10x 2.5″ hot-swap (NVMe/SAS/SATA)

• Networking: 4x 200G ConnectX-7 + OCP 3.0

• Power: 3200W Titanium (N+N redundant)

This system is particularly suited for:

• AI research facilities

• Hyperscale cloud providers

• Government supercomputing centers

• Enterprise AI labs