OIF is driving forward the next wave of innovation in AI and high-speed data transmission with the launch of three groundbreaking projects aimed at meeting the industry’s most pressing challenges. These initiatives, focused on enhancing energy efficiency, scaling AI capabilities, and pushing data transmission to new speeds, were unveiled at OIF’s Q3 2024 Technical and MA&E Committees Meeting in Ottawa, Canada.
These new projects, which emerged from the recent meeting, underscore OIF’s commitment to developing interoperable, high-efficiency solutions for the future of communication systems.
• EEI Compute Optics Interface (COI) for AI Scale-up: This project aims to develop energy-efficient, low-latency photonic interfaces to support the scale-up of AI and machine learning workloads. The focus is on creating standardized, interoperable solutions for local accelerator connectivity, such as PCIe, NVLink, and UALink. This project builds on existing EEI work and seeks to provide scalable, high-performance compute optical interfaces that ensure compatibility across diverse AI platforms.
• EEI-224G-RTLR Project: The EEI-224G-RTLR project is centered on developing an Implementation Agreement (IA) for 224G Retimed Tx Linear Rx (RTLR) optical interfaces. This project is critical for reducing power consumption and latency in next-generation data transmission systems. The RTLR approach eliminates the need for Rx DSP, resulting in significant cost and power savings—about half of what can be achieved with fully retimed pluggable optics. RTLR also enables true plug-and-play interoperability by decoupling optical and copper links, unlike traditional Linear Pluggable Optics (LPO).
• CEI-448G Framework Project: The CEI-448G Framework project is focused on defining the next electrical data rate beyond 224Gbps, targeting 448Gbps. This initiative aims to address the challenges of achieving this high-speed data rate by exploring key factors such as interface support, modulation schemes, and test methodologies. Input from end-users, equipment developers, and industry experts will be compiled into a technical white paper, laying the groundwork for future projects that will develop specific standards and architectures for 448Gbps data rates.
“Our in-person meetings provide a critical platform for collaboration and progress, allowing us to push the boundaries of innovation and address the evolving needs of the networking and AI/Machine Learning (ML) industries,” said Karl Bois, OIF Technical Committee Chair and NVIDIA.
