Intel remains a key driver to the virtualized radio access network (vRAN) market, having recently signed supply deals with large vendors to support their ongoing work and rolling out new updates that could reduce the need for custom silicon.

Sachin Katti, SVP and GM at Intel’s Network and Edge Group, explained that Intel’s recent deals with Samsung and Ericsson highlight the chip giants' ongoing work in helping telecom equipment vendors bolster their RAN equipment to power demanding 5G networks.

The Samsung deal has that vendor integrating its latest vRAN product onto Intel’s Xeon processor and vRAN Boost platforms in a push to further accelerate the performance of 5G RAN deployments. Specifically, Samsung is porting its vRAN 3.0 software onto Intel’s 4th-Gen Xeon Scalable processors that have the vRAN Boost accelerator platform secret sauce.

“The collaboration is about both current and future products that really help them scale faster on the Xeon 4th generation with vRAN Boost. That’s current.” Katti said of that deal. “But we are also working with them on future-generation products. We will be delivering a class of SOCs [system-on-chips] that is optimized for cloud RAN workloads every couple of years, roughly. We want to make sure that they have a very smooth transition and enjoy the performance.”

The Ericsson agreement is more bespoke, with Intel using its latest 18A process technology to manufacture custom 5G system-on-chip products for Ericsson. These will be used in Ericsson’s future 5G infrastructure.

“Custom silicon is a longer timescale because we are designing it with them and then doing the manufacturing, so that is the future part of that deal,” Katti said. “But the work starts now and because it's custom for them they are heavily involved in the design of those chips with us and for their own products.”

That customized nature remains important for vendors as they deal with increased performance demands, with Katti explaining that the decision to go with common-off-the-shelf (COTS) hardware or a customized solution is a matter of balancing tradeoffs.

“You can have different design points where instead of it being software that is portable across different pieces of silicon you could bake some of that software functionality into the silicon itself,” Katti said. “That often is a tradeoff between performance and power consumption and, obviously, flexibility in being able to do upgrades and change things after deployment. So it's a different design point in that sense.”

That potential performance tradeoff is significant. A recent Analysys Mason survey found 40% of operators looking to use vRAN “consider potential performance trade-offs to be one of the top barriers to deployment.”

“We believe eventually that these general-purpose platforms will just get good enough that you won't have to do this kind of strong customization, but there's always going to be benefits and trade-offs in between these different architectures and we remain open,” Katti said. “We work with the ecosystem in providing both kinds of approaches.”

This trade-off is becoming more important with the growing roll out of multiple-input, multiple-output (MIMO) antenna technology, which can embed dozens of smaller, smart antennas within a traditional antenna structure at a cell site. Those individual antennas need to be managed and controlled in real time, which can tax traditional processing capabilities.

“We expect that workloads will continue to emerge that will want some customization ability,” Katti said. “If you have enough scale in your world. If you are selling enough of your pieces of software or systems then it actually might make sense for you to do some custom. But if you do not have that scale or it's a smaller application, then it doesn't make sense because you do not have the scale needed to justify that semicustom environment.”

Katti also sees continued momentum behind the nascent Multi-G Initiative that is looking to open the RAN interface down to the layer-one (L1) silicon stack of a telecommunications network. Intel was one of the initiative's founding members alongside Cohere, Juniper Networks, Mavenir and VMware when it launched earlier this year. It also boasted carrier support from Bell Canada, Telstra and Vodafone.

The work is focused on disaggregating RAN intelligence and scheduling functions that can support updates to the underlying silicon architecture, Intel’s FlexRAN in this initial instance. This will allow that architecture to support higher capacity and software-defined deployments for legacy 4G LTE, current 5G and future next-generation wireless standards.

Katti explained that the work is focused on providing operators with more flexibility around their vRAN technology deployment choices.

“You cannot have the layer-one be baked into hardware like these inline accelerators because with this kind of ability to run even the layer one as a piece of glorified software, you can actually evolve a software architecture where you can support multiple generations of RAN protocols – 4G, 5G, whatever 6G becomes – into a single software framework. And to us this is super important because it gives our customers leverage,” Katti said.