SynaXG announced technical milestones in artificial intelligence-native radio access networks by demonstrating concurrent operation of 5G FR1, 5G FR2 and artificial intelligence workloads on the Nvidia artificial intelligence-RAN platform with real-time, policy-driven GPU orchestration. The company reported that software-defined, artificial intelligence-native RAN can deliver carrier-grade performance and continuous 24×7 operation on shared accelerated computing infrastructure, positioning the system as ready for commercial deployment. The platform is built on Nvidia artificial intelligence Aerial technologies, enabling a software-defined architecture that supports deterministic network performance and artificial intelligence processing on the same GPU infrastructure.
In performance testing, SynaXG achieved carrier-grade 5G FR1 performance on a single Nvidia GH200 platform. The system simultaneously operated 20 x 100MHz 5G NR cells, delivering more than 36Gbps of aggregated throughput with sub-10 millisecond latency, while supporting up to 1,200 connected user equipment per cell. The company stated that these results validate that a fully software-defined artificial intelligence-RAN on accelerated infrastructure can match best-in-class commercial 5G deployments while running artificial intelligence workloads concurrently without compromising network key performance indicators. SynaXG also reported the world’s first carrier-grade FR2 virtualized RAN running alongside FR1 and artificial intelligence workloads on a shared GPU platform, achieving high throughput with end-to-end latency as low as 5 milliseconds to meet dense urban and enterprise requirements.
SynaXG’s real-time orchestration software dynamically allocates and switches Multi-Instance GPU partitions based on live RAN and artificial intelligence key performance indicators, continuously rebalancing GPU resources across FR1 coverage, FR2 capacity and artificial intelligence workloads without service interruption to maintain deterministic RAN performance and maximize infrastructure use. The artificial intelligence-RAN system operated continuously, 24 hours a day, 7 days a week, under sustained load, validating stability, scalability and operational resilience for carrier environments and confirming that Nvidia GPU-accelerated artificial intelligence-RAN can maintain predictable performance under varying traffic. Looking ahead, SynaXG is extending its platform to additional Nvidia systems such as Nvidia DGX Spark to support a write once, run anywhere approach across Nvidia CUDA-based platforms, while advancing artificial intelligence-for-RAN capabilities for optimization, automation, spectral and energy efficiency and other features in collaboration with ecosystem partners.
