NOVA-AS-B200-LC B200 Liquid-Cooled AI Server Background
AI SERVERS

OPTA NOVA-AS-B200-LC B200 Liquid-Cooled AI Server

A high-density, energy-efficient liquid-cooled compute server for mid-to-large AI model training and high-density compute cluster scenarios.

Overview

The NOVA-AS-B200-LC is a high-density, energy-efficient liquid-cooled compute server designed for mid-to-large AI model training and high-density compute cluster scenarios. Featuring dedicated DLC cold plate cooling technology, it efficiently captures operational heat, significantly reducing cooling energy consumption and data center noise while improving compute equipment stability and service life. With ultra-large GPU memory and high-speed interconnect architecture, it requires deployment with a standardized liquid-cooled rack integration system — serving as a core compute node for efficient green AI clusters.

Key Specifications

Model:NOVA-AS-B200-LC
Processor:Dual Intel Xeon 6700 series (with P-cores)
GPU Configuration:8× Nvidia B200 GPUs, 1.44TB ultra-large GPU memory
Cooling Technology:DLC cold plate cooling, max configuration heat capture rate 87%
Memory Support:32 DIMM slots, supports RDIMM ECC DDR5 memory
Network & Management:2× 10GbE ports + dedicated BMC/IPMI management interface
Power Supply:4× 6600W titanium-grade redundant PSUs
Dimensions:448 mm (W) × 911.4 mm (D) × 174 mm (H)
Deployment Requirements:Requires standardized liquid-cooled rack integration system
NOVA-AS-B200-LC B200 Liquid-Cooled AI Server

Key Features

DLC Cold Plate Liquid Cooling

Dedicated DLC cold plate cooling with up to 87% heat capture rate, significantly reducing cooling energy and data center noise.

8× B200 Ultra-Large GPU Memory

8× Nvidia B200 GPUs with 1.44TB ultra-large GPU memory and high-speed interconnect for mid-to-large AI model training.

Titanium-Grade Redundant PSUs

4× 6600W titanium-grade redundant power supplies for stable operation in high-density compute scenarios.

Standardized Liquid-Cooled Cluster Deployment

Requires standardized liquid-cooled rack integration system — a core compute node for efficient green AI clusters.