NVIDIA launches next-generation X800 series network switches supporting 800Gb/s ultra-high-end end-to-end throughput

According to IT House reports, at today’s GTC Developer Conference, NVIDIA launched a new generation of X800 network switch platform. These platforms are the first in the industry to achieve 800Gb/s end-to-end throughput, further pushing the limits of network performance for artificial intelligence and computing workloads.

Specifically, the X800 series includes two platforms: Quantum-X800 for InfiniBand and Spectrum-X800 for standard Ethernet.

The Quantum-X800 platform is designed to provide the highest performance for artificial intelligence infrastructure, including the NVIDIA Quantum Q3400 switch (IT House Note: can provide 144 800Gbps ports) and the ConnectX-8 SuperNIC device-side smart network card.

Compared with the previous generation product, the total bandwidth of Quantum-X800 has increased by 5 times, and the in-network computing power using NVIDIA SHARPv4x protocol has also increased by 9 times, reaching 14.4 Tflops.

Spectrum-X800, on the other hand, targets artificial intelligence, cloud and enterprise infrastructure scenarios, and includes Spectrum SN5600 switches that provide 64 800Gbps ports and BlueField -3 SuperNIC device-side high-performance network cards.

NVIDIA claims that Spectrum-X800 accelerates the running of artificial intelligence workloads and is designed for cloud computing multi-tenant environments, ensuring performance isolation between each tenant's AI workloads to maintain optimal and consistent performance performance level, improving cloud service quality and customer satisfaction.

NVIDIA also provides a complete set of software support that can improve the communication efficiency of GPU parallel computing and facilitate customers to optimize the performance of large models with trillions of parameters.

The above is the detailed content of NVIDIA launches next-generation X800 series network switches supporting 800Gb/s ultra-high-end end-to-end throughput. For more information, please follow other related articles on the PHP Chinese website!