When AI workloads run alongside live 5G radio cells, shared GPU platforms begin handling both network processing and machine learning tasks in the same system.
SynaXG has demonstrated concurrent operation of 5G radio access network functions and artificial intelligence workloads on a shared accelerated computing platform built on NVIDIA’s AI Aerial infrastructure. The demonstration illustrates how software defined architectures can support both radio network processing and AI applications within a single system environment.
Radio access networks have traditionally relied on specialised hardware to process wireless signals and maintain connectivity between base stations and user devices. As 5G networks expand in scale and complexity, operators are exploring software defined and virtualised approaches that allow network functions to run on accelerated computing platforms while maintaining performance requirements such as throughput, latency and reliability.
In the demonstration, the system operated multiple 5G frequency ranges alongside AI processing tasks on a single platform. For sub 6 GHz spectrum, the setup supported 20 cells operating at 100 MHz bandwidth each, delivering aggregated throughput exceeding 36 gigabits per second with latency below 10 milliseconds while supporting up to 1,200 connected devices per cell.
The platform also ran millimetre wave radio access network functions alongside these workloads. In this configuration, end to end latency reached approximately 5 milliseconds while maintaining simultaneous operation with the lower frequency cells and AI processing tasks.
Resource allocation across the system was managed through real time orchestration software that dynamically adjusted GPU partitions based on network and AI performance indicators. By redistributing computing resources between coverage focused FR1 cells, capacity oriented FR2 cells and AI workloads, the system maintained continuous operation without interrupting network services. The architecture operated under sustained load in continuous operation, reflecting conditions typically required for carrier environments.
Xin Huang, CEO of SynaXG, says “With the recent industry-leading breakthroughs, SynaXG has demonstrated that AI-RAN can deliver carrier-grade FR1 and FR2 performance with continuous 24×7 operation on shared NVIDIA AI infrastructure.”
Soma Velayutham, VP of AI and Telecoms, NVIDIA, says, “Software-defined architecture is key to the next generation of wireless networks. SynaXG’s benchmark 5G performance on NVIDIA AI-RAN platform proves that operators and enterprises can achieve the flexibility and agility of cloud-native computing while maintaining carrier-grade throughput and performance per watt, required for commercial 5G services.”
