NVIDIA Launches Accelerated Ethernet Platform for Generative AI

Author photo: Patrick Arnold
ByPatrick Arnold
Company and Product News


NVIDIA has announced NVIDIA Spectrum-X, an accelerated Ethernet platform designed to improve the performance and efficiency of Ethernet-based AI clouds. NVIDIA claims the platform, powered by the coupling of NVIDIA Accelerated Ethernet PlatformSpectrum-4 Ethernet switches and NVIDIA Bluefield-3 data processing units, can enhance AI performance and power efficiency as well as provide predictable performance in multi-tenant environments. Spectrum-X also includes acceleration software and software development kits that allow developers to build software-defined, cloud-native AI applications. It uses fully standards-based Ethernet and is interoperable with Ethernet-based stacks.

The delivery of end-to-end capabilities reduces run-times of massive transformer-based generative AI models, enabling network engineers, data scientists, and cloud service providers to make more agile decisions. The platform offers performance isolation and AI performance visibility features that can identify bottlenecks.

As a blueprint and testbed for Spectrum-X reference designs, NVIDIA is building Israel-1, a hyperscale generative AI supercomputer to be deployed in its Israeli data center on Dell PowerEdge XE9680 servers based on the NVIDIA HGX H100 eight-GPU platform, BlueField-3 DPUs and Spectrum-4 switches.

Emerging technologies, such as generative AI, are forcing enterprises to push the boundaries of data center performance. With Spectrum-X, NVIDIA seeks to remove barriers for next-generation AI workloads and usher in new opportunities to transform industries.

Acceleration software driving Spectrum-X includes NVIDIA SDKs, such as Cumulus Linux, pure SONiC, and NetQ. It also includes the NVIDIA DOCA software framework, which is at the heart of BlueField DPUs.

Engage with ARC Advisory Group

Representative End User Clients
Representative Automation Clients
Representative Software Clients