SYS-822GS-NB3RT 8U HGX B300 NVL8

🔧 High-Performance AI Server Configuration

Model: 8U X14 HGX 8GPU | Chassis: Supermicro CSE-GP807TS

🧠 Compute Platform

Motherboard

• Model: Supermicro X14DBG-LC
• Form Factor: Proprietary for 8U GPU systems
• Chipset: Intel Eagle Stream (LGA4710)
• Expansion: Supports multiple PCIe Gen5 slots for GPUs, DPUs, and storage

Processor

• CPU: 2x Intel Xeon GNR-SP 6767P (Granite Rapids-SP)
  • Cores/Threads: 64 cores / 128 threads per CPU
  • Base / Turbo Frequency: 2.4 GHz / 2.8 GHz
  • TDP: 350W
  • Cache: 336MB L3
  • Memory Support: DDR5, up to 6400 MT/s
  • Use Case: Ideal for AI training, inference, and high-throughput data processing

🧬 Memory

• Total Capacity: 128GB DDR5 ECC RDIMM
• Module: 1x 128GB 2Rx4 (32Gb) ECC Registered DIMM
• Speed: 6400 MT/s
• Error Correction: ECC for mission-critical reliability
• Scalability: Supports multi-TB configurations for memory-intensive workloads

💾 Storage

Primary NVMe (High-Speed Data Ingestion)

• Model: Samsung PM9D3a
• Capacity: 7.68TB
• Interface: PCIe Gen5 x4
• Form Factor: E1.S 9.5mm
• Endurance: 1 DWPD
• Security: SED (Self-Encrypting Drive)

Secondary NVMe (Boot/OS/Application)

• Model: Samsung PM9A3
• Capacity: 1.92TB
• Interface: PCIe Gen4 x4
• Form Factor: M.2 22x110mm
• Endurance: 1 DWPD
• Warranty: 5 Years

🎮 GPU & AI Acceleration

GPU Baseboard

• Model: NVIDIA HGX B300 NVL8 (Blackwell Architecture)
• GPU Count: 8x NVIDIA Blackwell GPUs
• Interconnect: NVLink 5.0 for ultra-fast GPU-to-GPU communication
• Cooling: Active Cooling (AC)
• Use Case: LLM training, generative AI, scientific computing

DPU (Data Processing Unit)

• Model: NVIDIA BlueField-3 B3140H E-series
• Form Factor: HHHL (Half-Height, Half-Length)
• Networking: 400GbE
• Functionality: Offloads networking, storage, and security tasks from CPU

🛡️ Support & Warranty

• Standard 3-Year Support for ADPTR-BF-SW

🧩 Assembly & Integration

• Assembly Fee: Included
• Socket Carrier: E2 E2A, LGA4710, GNR-SP XCC, DG1.0, Cool Gray, Halogen-Free
• Thermal Design: Optimized for high-density GPU workloads

🧠 Ideal For

• Training large language models (LLMs)
• Generative AI and deep learning
• Scientific simulations and HPC
• Real-time inference at scale
• Data center AI infrastructure

NTS AI Software Stacks

ForgeKit: Maximum Control for Experts
Ideal for: Advanced researchers, HPC environments, secure and regulated sectors

• Secure Minimalism: Lean base images containing essential GPU drivers and CUDA Dev Toolkit to reduce risk and enhance security.
• High Customizability: Customize environments with bespoke compilation options, experimental libraries, and optimized NCCL builds for maximum multi-GPU efficiency.
• Compliance Simplified: Designed to streamline audits, ForgeKit excels in regulated sectors including defense, healthcare, and secure cloud infrastructures.