SANTA CLARA — NVIDIA has introduced the Quantum-X800 InfiniBand and Spectrum-X800 Ethernet, setting a new benchmark for networking performance tailored to the demands of AI-dedicated infrastructure. Announced at the prestigious GTC event, these cutting-edge platforms represent a significant leap forward in computing capabilities, underlining NVIDIA’s commitment to empowering the evolution of AI-driven technologies.
The Quantum-X800 and Spectrum-X800 platforms, boasting an unprecedented end-to-end throughput of 800Gb/s, redefine the boundaries of networking performance for AI, cloud, data processing, and HPC applications. With advanced software acceleration features, these platforms accelerate AI workloads across diverse data center environments, including those leveraging the innovative Blackwell architecture-based product lineup.
Gilad Shainer, Senior Vice President of Networking at NVIDIA, emphasized the pivotal role of NVIDIA Networking in driving the scalability of AI supercomputing infrastructure. The X800 switches serve as the backbone of NVIDIA’s AI infrastructure, enabling the realization of trillion-parameter-scale generative AI crucial for pioneering AI architectures.
Early adopters of Quantum InfiniBand and Spectrum-X Ethernet, including industry giants like Microsoft Azure and Oracle Cloud Infrastructure, underscore the transformative potential of NVIDIA’s networking solutions in powering high-performance AI engines.
The Quantum-X800 platform, comprising the Quantum Q3400 switch and the ConnectX-8 SuperNIC™, delivers unmatched performance for AI-dedicated infrastructure, achieving a remarkable 5x increase in bandwidth capacity and a 9x boost in In-Network Computing compared to previous generations.
On the other hand, the Spectrum-X800 platform, featuring the Spectrum SN5600 switch and the BlueField-3 SuperNIC, caters to the specific networking requirements of AI cloud and enterprise infrastructure, facilitating faster processing, analysis, and execution of AI workloads.
NVIDIA’s comprehensive software support, including the Collective Communications Library (NCCL), extends GPU parallel computing tasks to the Quantum-X800 network fabric, enhancing performance for large model training and generative AI.
With Quantum-X800 and Spectrum-X800 set to be available from a diverse array of infrastructure and system vendors globally, including leading names like Dell Technologies, Hewlett Packard Enterprise, and Lenovo, NVIDIA is poised to drive widespread adoption of these revolutionary networking platforms, ushering in a new era of AI infrastructure scalability and efficiency.
