No image available
MPN: AICL-G-NVI-1
NVIDIA Cluster Node Essential G-1
Highlights
•Scale: Rack-scale building block
•GPU Density: Up to 72 GPUs per rack
•Interconnect: NVLink Switch System
•Cooling: Direct Liquid Cooling
•Networking: 800G InfiniBand spine
Starting at :$9,500.00
Rack-scale infrastructure built for the largest generative AI training jobs. Pre-integrated, pre-cabled, and pre-burned-in so your team can move from delivery to training in days, not months.
The Data Center as the New Unit of Compute
Rack-scale infrastructure built for the largest generative AI training jobs. Pre-integrated, pre-cabled, and pre-burned-in so your team can move from delivery to training in days, not months.
NVLink Switch System unifies 72 GPUs into a single accelerator. DLC eliminates fan noise and improves PUE. Tested at scale with leading LLM frameworks.
| Specification | Value |
|---|---|
| Rack GPUs | 72 × NVIDIA Blackwell |
| NVLink Bandwidth | 130 TB/s all-to-all |
| Compute Nodes | 18 × dual-CPU compute trays |
| Switch Trays | 9 × NVLink Switch trays |
| Cooling | Direct-to-chip liquid cooling (DLC) |
| Power Distribution | 120 kW per rack, busbar fed |
| Networking | 800G InfiniBand + 200G management |
| Deployment | Pre-integrated, pre-tested rack-scale solution |