Marvell Showcases 200G/Lane Partner Demonstrations at DesignCon, Extending Connectivity Leadership

Amphenol, Keysight Technologies, Molex and TE Connectivity showcase 200G/lane copper links with Marvell DSPs for increasing data center bandwidth

Marvell Technology, Inc., a leader in data infrastructure semiconductor solutions, and leading cabling and system partners Amphenol, Keysight Technologies, Molex and TE Connectivity (TE) this week are demonstrating copper interconnects powered by Marvell 224G long-reach (LR) DSP SerDes operating at 200 Gbps per lane or faster, a critical milestone in the race to scale accelerated infrastructure for AI and cloud workloads.

The demonstrations are taking place during DesignCon at the Santa Clara Convention Center January 30 through February 1, 2024 in Santa Clara, California.

Cloud service providers are undertaking wide-ranging infrastructure upgrades to accommodate the insatiable demand for generative AI and other services. Networking bandwidth in the cloud is growing at over 40% per year while bandwidth dedicated to AI is growing at over 100%.¹ Without faster, higher capacity networks, workloads will require more time, energy, and money to complete, potentially upending the economic potential of AI and cloud services.

Recommended AI News: Cisco Collaborates With Microsoft and Samsung to Deliver Superior Meeting Room Experiences

Electrical copper cables are used for short reach (0-5m) server to top-of-rack switch connectivity and for certain cloud-optimized intra-rack interconnects in data centers. Boosting baseline I/O lane speed to 200G/lane represents a 2x increase in bandwidth over current leading-edge 100G/lane systems. Marvell 200G/lane LR DSP SerDes are expected to be incorporated into a wide range of networking platforms that will extend the reach of copper links with cabled-backplane, cabled-host connections and other cabling solutions. Marvell 200G/lane LR DSP SerDes will also extend copper connectivity bandwidth and enable 1.6T Active Electrical Cables (AECs).

“Driven by the market’s need for increased speed, Molex is collaborating with Marvell to deliver an industry-leading, comprehensive 224G connector, cable, and backplane portfolio that ensures optimal channel performance for AI, machine learning (ML) and 1.6T high-speed applications,” said Vivek Shah, director, New Product Development, Molex.

“Keysight is working with its partners and standard bodies to enable the next generation high-performance computing infrastructure to cope with the rapidly expanding computing and networking needs to support the upcoming AI applications,” said Dr. Joachim Peerlings, vice president and general manager of Keysight’s Network and Data Center Solutions Group.

Recommended AI News: Kubecost Launches Kubecost 2.0 with Kubernetes Network Monitoring, Cost Forecasting, and More

“Active electrical technology extends the life of copper in data centers,” said Alan Weckel, co-founder of 650 Group. “It will quickly become one of the standard building blocks for data centers. AEC silicon revenue is expected to grow at 64% per year and achieve sales of $1 billion by 2028 with chip units for powering AEC devices reaching nearly 40 million per year.”

“Virtually every class of connection in the cloud—from connections between distant data centers to connections between components within a chip package—will be transformed over the next several years. Active electrical cable (AEC) connectivity will be a critical component for short reach interconnects at 200G,” said Venu Balasubramonian, vice president of product marketing for the High Speed Connectivity Group at Marvell. “We are actively collaborating with our ecosystem partners and our hyperscale customers to deploy 200G/lane copper interconnect technologies to enable the scaling of bandwidth needed to keep up with the pace of AI innovation.”

Recommended AI News: Sema4.ai Raises $30.5 Million to Bring Open Source Powered AI to Mission-Critical Enterprise Work

[To share your insights with us as part of editorial or sponsored content, please write to sghosh@martechseries.com]