OPTA™ NOVA-AS-B300-AF B300 Air-Cooled AI Server
A flagship air-cooled AI compute server for ultra-large model training, high-end AI inference and massive data processing workloads.
Overview
The NOVA-AS-B300-AF is a flagship air-cooled AI compute server designed for ultra-large model training, high-end AI inference and massive data processing scenarios. Featuring the NVIDIA HGX B300 flagship GPU cluster with fifth-generation NVLink high-speed interconnect for ultra-high bandwidth compute synergy, paired with 800Gb/s ultra-high-speed network cards for significantly improved compute transmission and operation efficiency. A high-performance benchmark under air-cooled architecture, ideal for core compute deployment in high-end enterprises and AI R&D organizations.
Key Specifications
Key Features
HGX B300 Flagship GPU Cluster
NVIDIA HGX B300 8-GPU cluster with 5th-gen NVLink 1.8TB/s interconnect for ultra-high bandwidth compute synergy.
2.3TB HBM3e Ultra-Large Memory
2.3TB HBM3e ultra-large GPU memory per system for ultra-large model training and high-end AI inference.
800Gb/s Ultra-High-Speed Network
8× NVIDIA ConnectX-8 SuperNICs with up to 800Gb/s bandwidth for significantly improved compute transmission and efficiency.
Air-Cooled Performance Benchmark
6× (3+3) 6.6kW titanium-grade redundant PSUs — a high-performance benchmark under air-cooled architecture for enterprise and AI R&D deployment.