AI in data centers is rapidly evolving towards ever larger deep learning models with trillions of parameters. Therefore, next generation AI systems must evolve to utilize dense electrical and optical interfaces. Without interfaces offering greater energy efficiency and lower latency, these high-performance next generation systems become impractical to build. The OIF’s Energy Efficient Interfaces track is addressing these challenges.

Energy Efficient Interfaces Framework Project: The Energy Efficient Interfaces Framework project is studying end-user requirements and will identify energy efficient, low latency electrical and optical interfaces necessary to support the future application requirements for AI in the data center. These interfaces include electrical and optical interfaces for co-packaged, near-packaged, and for pluggables and includes retimed, transmit retimed and linear interfaces.

Retimed Tx Linear Rx (RTLR) Project: As a direct off shoot of the EEI Framework Project, OIF members voted to start the RTLR project. The RTLR project addresses energy efficiency and low latency requirements of pluggable optics for Ethernet and AI/ML at up to 200G/lane while achieving full electrical and optical plug-and-play. Specifications will be developed for 200G/lane (DRn and 800G-FR4-500 over 500m) as well as for 100G/lane (30m MMF, DRn, 400G FR4).

For more information, contact Jeff Hutchins, Physical and Link Layer Working Group Management Co-Vice Chair.

Press Releases


  • Speaking

      • OFC 2024: “Energy Efficient Interfaces – Reining in Power Consumption Trends for Next-Generation Optical Networking”
        Thursday, March 28, 2024
        Moderator: Jeff Hutchins, OIF PLL WG EEI Vice Chair and Board Member, Ranovus
        Panelists: Craig Thompson, NVIDIA; Yi Tang, OIF PLL WG Electrical Vice Chair, Cisco; Nathan Tracy, OIF President, TE Connectivity

    For information on other OIF current projects, please see OIF Current Work.

    For public OIF Implementation Agreements, please see Implementation Agreements (IAs).