The need for scalable compute in a 6G network

For 6G and AI, where large amounts of data must be processed, energy-efficient and scalable computing is essential. But this can only be achieved by having more silicon options for the RAN.​

​For vendors one of the barriers to adopting alternative silicon architectures has been the cost and complexity of adapting their software to different platforms. This can require maintaining multiple code bases, making migrating to alternative silicon complex.

Equally for MNOs, there is uncertainty over whether the benefits of moving to a new architecture, such as lower energy consumption, will offset the cost and complexity of a migration.

Increasing silicon options for the RAN

Together with Arm, we set about addressing this challenge. As part of UK Government’s Future RAN programme, Arm worked with partners, including CC, to abstract Single Instruction, Multiple Data (SIMD) functions (a type of parallel processing) across Arm and other silicon architectures without compromising performance.

CC worked with Arm to create a single codebase (using Google Highway) that is optimised to run efficiently on both Arm and other silicon architectures. This proved the value of a flexible abstracted model for service delivery.

This work is now being developed further by existing vendors to test the ability to port their software seamlessly across silicon options.

Increasing power efficiency and/or performance as needed

This is a powerful example of what is now possible with emerging tools and platforms, allowing MNOs to use a single codebase on a range of silicon architectures and to streamline migration to Arm based compute and other silicon.

By opening up the RAN market to new silicon vendors, this work will support moves to introduce competition and alternative, more power-efficient networks. It will also speed up the market adoption of new, more flexible RAN architectures, allowing operators to map compute to their needs and prepare for growth in data adoption of AI and the evolution to future 6G services.

Evolution of the telecoms network

Interest in Virtualized Radio Access Networks (vRANs), Open Radio Access Networks (O-RAN), AI and more efficient energy usage are all challenging the status quo. With the disaggregation of the network Base Band Unit (BBU) across Radio Units (RUs), Distributed Units (DUs) and Central Units (CUs), and the growth of the Cloud, vendors and operators are empowered to think differently about where the processing for different functions in the networks could and should occur.

Clear choices now exist between specialist highly tuned silicon, such as ASICs or FPGAs, for time-critical high-performance applications, and the use of standard commercial off-the-shelf (COTs) servers and specialised GPUs for AI applications. The choice of computing platform depends on where this compute should happen in the network; and involves trade-offs between full Cloud deployments and more distributed compute depending on scale economies, latency of key functions, availability of fibre between network elements and, increasingly, energy consumption.

As the RAN evolves and the market prepares for later 5G releases, not to mention the introduction of more AI services in a 6G network, flexibility around silicon options will be increasingly important.

See the demo at Mobile World Congress (MWC) 2025

Arm will be showcasing a demo of this project on their booth at MWC – Hall 2 Stand 2I60. Visit the Arm booth to see the demo and explore impressive performance benchmarks.

Continuing the conversation…

If you are interested in learning more, please drop me an email. Alternatively, meet me at Mobile World Congress next month, where CC will be part of the Capgemini booth – Hall 2, stand 2K21.