Featured Story
Ericsson CEO champions his silicon over Nvidia chips in RAN #
Per Narvinger, Ericsson's new CEO, describes Ericsson's ASICs as an 'extraordinary asset' when asked about interest in Nvidia's GPUs.
T-Mobile takes strides in AI-RAN with the first call with Nokia and Nvidia, and performance gains with Ericsson's AI beamforming.
T-Mobile notched up AI-RAN milestones with both of its radio access network vendors Ericsson and Nokia this week, signaling the various ways it is adding intelligence to its wireless networks.
Nokia said that with Nvidia, T-Mobile completed the "world's first" AI-RAN voice call in a live commercial network.
Few details were provided other than that the call was made in the live production network at a site near the carrier's headquarters using Nokia's massive MIMO radio, anyRAN software and server infrastructure built on Nvidia's Aerial RAN computer (ARC) platform.
Nokia's Chief Technology Officer Pallavi Mahajan wrote in a blog post that the test is an important moment for its technology ambition to move from the lab and into the network, just one year since the Finnish vendor partnered with Nvidia. As the world's first "software-defined RAN call using accelerated computing," it is "a moment from which there is no looking back," she said.
There is bound to be some relief within the halls of Nokia, since T-Mobile CTO John Saw said he would be "really upset" if he did not have the vendor's AI-RAN system to test in the field within the year, speaking during Nokia's press conference at Mobile World Congress in March.
Nokia said the next phase for the field trials is to explore other spectrum bands as well as supporting AI inferencing workloads with RAN functions. "Each step brings us closer to understanding how AI-RAN can deliver meaningful benefits in real-world network environments and ultimately operate at scale," wrote Mahajan.
The news coincided with Nvidia's GTC event in Berlin, where there was a host of AI-RAN trial announcements with carriers, including BT, Deutsche Telekom and Elisa.
Just the day before, T-Mobile revealed a different AI achievement in its live 5G-Advanced network in New York City with Ericsson with measurable performance improvements.
In a large-scale commercial trial, T-Mobile is implementing Ericsson's AI-managed beamforming. Beamforming essentially does as it sounds – it narrows a wireless signal to focus more accurately on user devices using information about the radio environment through sounding reference signals (SRS).
AI-managed beamforming makes that process smarter. An AI model trained on T-Mobile's network data runs directly on Ericsson's RAN compute and silicon. This allows the basestation to predict where demand is and "dynamically" shift SRS resources where needed.
In the New York City sites where the feature is activated, T-Mobile clocked the best results during peak usage hours. The carrier found 16% improvement in downlink throughput; 12% better spectral efficiency; 10% increase in downlink data volume; and a 50% more advanced beamforming using full-band SRS.
"This is a wow moment," said Gabriel Brown, senior principal analyst at Omdia, a Light Reading sister company. "It shows AI models deployed directly in the basestation can give very measurable and significant performance gains, and that's with software updates to deployed equipment."
Further, this is the second AI model T-Mobile has deployed in Ericsson RAN this year. The carrier also ran "AI-native scheduler with link adaptation" in a large-scale trial, which showed 10% better spectral efficiency and a 15% increase in downlink throughput.
"These are initial deployments but there could be many more AI models still to come… We could almost start stacking the gains from each. We don't really know quite how many different models can be used simultaneously, or what the overall gain will be, but it's looking like significant," said Brown.
With T-Mobile's AI-RAN trials with both Nokia and Ericsson, the carrier is like a microcosm for the RAN platform strategy divergence between the two vendors. Nokia is on an Nvidia GPU-based platform path, while Ericsson is dedicated to its own silicon. T-Mobile is showing milestones with both.