NVIDIA Quantum-2 QM9700

Description

The NVIDIA Quantum-2 QM9700 is one of the most advanced InfiniBand switches on the market, designed to meet the needs of AI supercomputers, HPC clusters, and enterprise data centers. It combines unmatched bandwidth, ultra-low latency, and smart networking features to accelerate the most demanding workloads, from training large AI models to running complex scientific simulations.

Unlike traditional switches that only forward data, the QM9700 comes with in-network computing (SHARPv3), adaptive routing, congestion control, and even self-healing capabilities. This means your network not only delivers raw speed but also optimizes itself for efficiency, stability, and performance at scale.

Technical Specifications

Why QM9700 is the Right Choice

• High-speed networking: 64 ports of 400Gb/s provide a total of 51.2 Tb/s throughput for uncompromised performance.

• Ultra-low latency: Sub-microsecond response time ensures smooth communication between GPUs and compute nodes.

• AI-ready features: SHARPv3 reduces communication overhead, accelerating collective operations in AI and HPC workloads.

• Scalable management: Onboard Subnet Manager supports up to 2,000 nodes, making cluster expansion simple.

• Efficient and compact: 1U design with 720W typical power consumption makes it both space- and energy-efficient.

Perfect for AI & HPC at Scale

The QM9700 is designed to meet the rising demands of modern AI networking and HPC environments. As the backbone of DGX SuperPOD architectures, it is the ideal choice for organizations running large-scale AI training clusters, foundation models, generative AI workloads, and scientific simulations. With its unmatched speed, intelligent features, and reliability, the NVIDIA Quantum-2 QM9700 is more than just a switch—it is the engine driving the future of supercomputing.