Marvell is open sourcing software for building machine learning models that can be executed on OCTEON 10 processors, which are widely deployed in base stations around the world. The move paves the way for OEMs and operators, such as Nokia and Vodafone, to develop and implement ML/AI models using the network infrastructure.
Some highlights
- The OCTEON 10 processor family is purpose-built to serve as the primary processing unit in RUs, DUs, enterprise networking equipment, or as a data processing unit in clouds and AI clusters. It incorporates digital signal processors, Arm Neoverse CPUs, and optimized hardware accelerators like the inference processor and inline crypto engine.
- Using open-source ML frameworks such as TensorFlow, PyTorch, and ONNX, developers can create their own models for optimizing the performance of the RAN Distributed Unit (DU) and Radio Unit (RU). Via standard APIs, these models can be implemented on the OCTEON 10 ML/AI accelerator, providing OEMs and service providers with enhanced ability to optimize massive MIMO beam selection, improve channel estimation, identify security risks or address other tasks to improve network performance, efficiency and operating economics.
- The OCTEON 10 ML/AI accelerator features a scalable tiled architecture and best-in-class performance-power ratio at 8 trillion operations per second per watt. The new API access follows the same principles used in DPDK’s Ethernet Device framework and Crypto framework already supported by OCTEON 10 processors.
“We are proud to be the first vendor to incorporate AI/ML into our ReefShark SoCs and AirScale portfolio. This technology allows us to differentiate our solutions and offer the best spectral efficiency and cell edge performance in the industry. We are now at the forefront of bringing machine learning technology to our customers and exploring its enormous possibilities ahead of the 6G era,” said Mark Atkinson, Head of RAN at Nokia.
“The ever-increasing complexity of RAN makes AI/ML the ideal technology to efficiently solve challenges around radio resource configuration, allocation, and optimization. Vodafone is a pioneer in this field, having deployed multiple uses of AI/ML in our network since 2017,” said Francisco Martin (Paco), Head of Open RAN at Vodafone.
“Our commitment to open-source integration is evident in the vast Marvell contribution to DPDK, OPI, and similar initiatives,” said Will Chu, senior vice president and general manager, Custom, Compute and Storage Group at Marvell. “Giving developers access to processor modules will enable industry innovations that build smarter, faster, and more secure RAN networks.”
