Description

Designed for NVIDIA H200 NVL GPUs, the 4‑Way NVLink Bridge (SKU S4A91C) provides a high-bandwidth, low-latency interconnect architecture, enabling seamless GPU-to-GPU communication among up to four GPUs. With total NVLink throughput of 1.8 TB/s and combined memory pooling of 564 GB HBM3e, this bridge triples the connectivity compared to 2‑way configurations—ideal for modern AI inference, large language model hosting, and HPC workloads. It bypasses PCIe Gen5 bottlenecks and accelerates parallel compute tasks by enabling shared memory and near-direct data exchange. Compatible with NVIDIA-Certified systems and scalable enterprise server designs, it’s essential for unlocking multi-GPU performance in data center and research environments.