Edgecore Unveils New 800G Ethernet Switch for ML and AI Workloads

Edgecore Networks

Edgecore Networks, a developer of open networking solutions, has introduced a switch called the DCS560 that is designed for 800 gigabits of traffic and may serve as an Ethernet-based fabric for artificial intelligence and machine learning workloads. The DCS560 is a 2RU system that has a throughput of 51.2 Tbps and contains 64x800G ports.

It is offered in two variations, each of which gives users the option of using either OSFP800 or QSFP-DD800 interfaces.

The recent interest in generative AI has led to a considerable increase in the size and network bandwidth requirements for AI and machine learning clusters. This is due to the significant growth in the number of compute nodes and accelerators. According to Edgecore, AI networks need to be built using high-capacity systems that have a flat design and are able to process massive volumes of data quickly and efficiently with a minimum of delay. Ethernet-based fabrics are becoming more popular as a means to cut the amount of time needed to do a project in order to satisfy the expectations of AI and ML workloads, added Edgecore.


A high-radix, deployment-friendly Ethernet-fabric is provided by Edgecore’s 51.2 Tbps Broadcom StrataXGS Tomahawk 5 series-based system with 64x800G ports. In order to ensure five-nine high availability and a broad climatic operating range for data center cloud applications, the 2RU system design would be both durable and packed into a small form factor with redundancy for both the power supply and the fan tray. According to Edgecore, the system provides great system quality and reliability while maintaining port assignment flexibility for customers by using a load-balanced port mapping design that does not employ flyover cables.

For flexible deployment, the platform is available with either OSFP800 or QSFP-DD800 interface choices. It also supports passive copper DAC on all ports and long-distance ZR+ optics. Each system would deliver high-radix connection to accelerators and computing nodes in a flat topology that would decrease both latency and the necessary amount of power. This would allow networks to be scaled-out in a sustainable manner.

“With the availability of Edgecore‘s 800G system, hyperscalers will want to take advantage of the increase in radix and throughput offered by the 800G AI fabric together with the reduced power consumption,” said Heimdall Siao, President of Edgecore. “The ground-breaking 800G system has a unique architecture that will bring a considerable boost in AI cluster performance and allow a cheaper total cost of ownership. Both of these benefits will be enabled. We are pleased to inform you that the DCS560 is now available, and we look forward to demonstrating the system at forthcoming events taking place all over the world.”