NVIDIA NVLink Bridge 4-Way Passive for Hopper H200 NVL GPUs 900-23946-0000-000

High-performance computing demands interconnect solutions that eliminate bottlenecks between GPUs. The NVIDIA NVLink Bridge 4-Way Passive delivers 448 GB/s of total bidirectional bandwidth, enabling two Hopper H200 NVL GPUs to communicate at unprecedented speeds for AI training, deep learning, and data center workloads.

Advantages

This bridge harnesses 4th Generation NVLink technology with a tool-less clip-on design, ensuring effortless installation and robust physical connectivity. Its passive cooling and aluminum alloy construction provide silent, reliable operation without additional power or thermal management. Each of the four links achieves 112 GB/s bidirectional throughput, dramatically reducing latency and accelerating multi-GPU scaling.

Suitable for

• AI researchers and data scientists running large-scale model training
• HPC clusters requiring low-latency GPU-to-GPU communication
• Enterprise data centers deploying NVIDIA Hopper H200 NVL GPU servers
• System integrators building high-density GPU compute nodes

Key specifications

• Device type: NVLink Bridge
• GPU compatibility: NVIDIA Hopper H200 NVL
• NVLink generation: 4th Generation
• Configuration: 4-Way (4 NVLink connections)
• Cooling type: Passive
• Maximum bidirectional bandwidth per link: 112 GB/s
• Total bidirectional bandwidth: 448 GB/s
• Interconnect technology: NVIDIA NVLink
• Form factor: GPU interconnect bridge
• Material: Aluminum alloy
• Color: Black/Silver
• Installation: Tool-less clip-on design
• Supported GPU count: Connects 2 GPUs
• Part number: 900-23946-0000-000

Recommendations

For maximum performance, pair this bridge with NVIDIA H200 NVL GPUs in a well-ventilated chassis. Ensure the server supports 4-way NVLink topology and verify GPU spacing before installation. No additional drivers are required—NVLink is natively supported by CUDA and NVIDIA drivers.