Cisco has expanded the options for customers looking to build AI clusters by adding support for Intel Gaudi 3 servers to its Nexus switch offering.
Specifically, Cisco said its 2U, Nexus 9364E-SG2 switches and OSFP-800G-DR8 transceivers are certified to support Intel Gaudi 3 AI accelerators in scale-out, high-performance configurations targeting LLM training, inference, and generative AI workloads.
Intel Gaudi 3 servers contain eight Gaudi 3 accelerators, Cisco stated in a white paper describing the integration: “Language models that can’t fit in or train fast enough using eight accelerators, can benefit by connecting multiple Gaudi 3 servers via an 800 Gigabit Ethernet (GbE) non-blocking network allowing RoCEv2 transport. A Gaudi 3 server provides 24 x 200 GbE scale-out or back-end interfaces exposed via six OSFP ports for an inter-Gaudi 3 network, also known as a scale-out network or a back-end network. These ports … are dedicated to interconnecting Gaudi 3 accelerators in other servers and are not connected to external networks. A Gaudi 3 server also provides additional interfaces for external, storage, and management connectivity.”
Cisco Nexus 9364E-SG2 switches provide 64 x 800 GbE or 128 x 400 GbE or 256 x 200 GbE or 512 x 100 GbE interfaces. These switches are available in two models. Cisco Nexus 9364E-SG2-O offers ports in OSFP form factor, and Cisco Nexus 9364ESG2-Q offers ports in QSFP-DD form factor. Except the port form factor, both switch models are similar, and both can be used for connecting Intel Gaudi 3 servers. These switches can be connected to the Intel Gaudi 3 servers using 4 x 200 GbE, Cisco stated.
As with any Nexus implementation, “Cisco Nexus Dashboard serves as the centralized management platform for AI fabrics, providing end-to-end RoCEv2 visibility and built-in templates for AI fabric provisioning,” stated Titus Shem, a product marketing manager on Cisco’s data center and networking team, and Paresh Gupta, a technical marketing engineer at Cisco, in a blog post about the Gaudi integration.
The idea is to offer customers a variety of networking and processor options for enterprise and hyperscaler customers looking to build out AI backbones and clusters.
Partnerships, validated designs strengthen Cisco offerings
Cisco’s AI offerings also include Nvidia technologies, such as Spectrum-X-based switches that are part of Cisco Secure AI Factory with Nvidia.
Cisco also works with AMD and its Instinct AI GPUs for networking and compute stack in large AI clusters. In addition, Cisco integrates AMD Pensando DPUs for advanced network and security acceleration.
“The journey to AI is not something that any single vendor or offering can give the customer everything they need. Instead, partnerships help provide more options and best-of-breed solutions. Customers want vendors to collaborate to create a better solution,” wrote Alan Weckel co-founder and analyst of the 650 Group, in blog about Cisco networking. “For example, Nvidia and Cisco’s partnerships allow for Cisco’s networking stack to integrate with Nvidia’s ASICs and GPUs to provide a lower TCO and an easier-to-manage solution. Similarly, Cisco and AMD collaborated on both the AI stack and a DPU to enable a Smart Switch (adding programmable data-plane functions), thereby enhancing the networking layer,” Weckel stated.
Enterprise customers require simplified network operations to scale, often seeking validated designs and architectures, wrote Weckel. “Cisco addresses this segment of the market with the Nexus 9000 and Nexus Dashboard, which allows for simplified network management. Telemetry-driven visibility and policy automation help teams correlate workload health with fabric behavior. Enterprise customers can benefit from adopting Smart switches, which enhance networks by adding embedded security capabilities.”
SOCIAL SHARE CARD GENERATOR