NX-21X208G – 2U 2-Node NVIDIA GH200 Superchip
Key Applications
- LLM training
- Generative AI and high performance computing workloads
- AI Inference and Machine Learning
- FP8 deep learning
- Molecular Dynamics Simulation
Outstanding Features
- 480 GB CPU LPDDR5X ECC Memory Per Node
- NVIDIA H100 GPU with 96 GB HBM3 VRAM Per Node
- 2 x PCI-E 5.0 x16 Dual Slot FHHL Slots Per Node
- 2 x PCI-E 5.0 x16 Single Slot FHHL Slots Per Node
- 1 x M.2 Slots(x4) Per Node
Specs
| Motherboard | |
|---|---|
| Processor Support | Each node (1 of 2): NVIDIA Grace Hopper Superchip |
| Chipset | NVIDIA Grace Hopper |
| System Memory (Max.) | 480 GB of LPDDR5X Per Node |
| Networking | Dual 10 GbE X550-T2 Per Node |
| PCIe Networking Options | Optional 1/10 GbE, 25 GbE, 40 GbE, 100 GbE Ports Available |
| VGA/BMC | ASPEED AST2600 BMC |
| Remote Management | Redfish API, IPMI 2.0; KVM with Dedicated LAN |
| Chassis | |
|---|---|
| Form Factor | 2U Rackmount |
| Dimensions(mm) | 850mm x 440mm x 87.5mm |
| Storage | |
|---|---|
| Onboard Storage Controller | Nvidia NVMe Controller |
| RAID Controller Options | Optional Hardware RAID or SAS HBA Available |
| Drive Bays | 8 x 2.5” Hot-Swap |
| Expandability | |
|---|---|
| Expansion Slots | 2 x PCI-E 5.0 x16 FHHL Dual Slot 2 x PCI-E 5.0 x16 FHHL Single Slot |
| Power | |
|---|---|
| Power Supply | 2 +1 3000W Redundant High-Efficiency Power Supplies with I2C & PMBus |
I/O Layout
