1. Product Overview & Key Highlights
- Ultra-Low Latency Interconnects: Sub-microsecond latency performance powering AI training clusters, High-Performance Computing (HPC), and enterprise cloud data centers.
- Dual Networking Fabrics: Available in Quantum InfiniBand (HDR 200G / NDR 400G / XDR 800G) and Spectrum Ethernet (10G to 400G Spectrum-2/3/4) architectures.
- In-Network Computing Acceleration: NVIDIA SHARP (Scalable Hierarchical Aggregation and Reduction Protocol) offloads collective communication operations directly to the switch fabric.
- Advanced Congestion Management: Hardware-accelerated telemetry, adaptive routing, and explicit congestion notification (ECN) to guarantee loss-free packet transmission.
- Flexible Network Operating Systems: Supported across MLNX-OS, Cumulus Linux, and SONiC open networking platforms for seamless orchestration and automation.

