• Intel’s new system-on-a-chip (SoC) features integrated accelerators, which help close the performance gap between virtual RAN (vRAN) and traditional RAN. Thus far, accelerators have been offered as separate hardware cards.
• Rakuten Symphony’s plans to offer a vDU based on the new SoC, following the availability of Qualcomm’s new accelerator and Juniper’s recent move to give away its RAN Intelligent Controller, show vRAN momentum building.
Recently, vRAN vendor Rakuten Symphony revealed plans to produce a virtual distributed baseband unit (vDU) based on new Intel SoC, due in 2023, whose accelerator is integrated with the CPU rather than being offered as a separate hardware card. This represents a departure from the status quo (accelerators as separate hardware cards), but the two vendors say it also addresses a key obstacle that has been holding vRAN back.
In general, vRAN adoption among operators has thus far been impeded by the fact that – especially when it comes to the massive MIMO radios that are so important to today’s 5G networks – vRAN doesn’t yet deliver the same performance as traditional RAN for a given amount of power consumption.
The industry has tried to narrow this performance gap with the use of supplemental hardware accelerators that take on some of the fundamental Layer 1 processing work. By integrating the accelerator functionality into the CPU SoC, Intel has notably altered vRAN architecture, reducing complexity by eliminating a network element. But more importantly, according to Rakuten Symphony CEO Tareq Amin, the new SoC will also double capacity relative to previous designs, allowing operators to serve twice the amount of spectrum or twice the number of subscribers, for example, by using the same level of power consumption. With that advancement, Intel believes it can double the performance per watt with each generation of vRAN silicon, thus tackling a primary impediment to vRAN’s broader adoption. When will that performance per watt match or exceed that of traditional RAN? Intel will only say that it is investing in five process nodes in the next four years and that that tipping point is “not very far away.”
For Intel, this move is the latest salvo against key rival Qualcomm, which recently made its own vRAN accelerator card available. The two chip giants’ opposing architecture approaches for vRAN have thus far fragmented the space: Qualcomm has favored ‘inline’ accelerators, which take on more Layer 1 functions, while Intel has offered ‘lookaside’ accelerators, which rely on the vDU CPU to do more of the work. Intel’s new, more integrated approach could allow the vendor to argue that Qualcomm’s approach is too dependent on extra hardware.
For Rakuten Symphony, the Intel news comes a month after the availability of the previous generation of its own Intel-based vDUs, of which Symphony plans to deploy 30,000 across more than 23,000 sites by Q3 2023 for parent company Rakuten Mobile’s Japanese network.
And just recently, Symphony announced plans to make Symworld (i.e., the vendor’s app store for Open RAN network management functionality) available through Juniper Networks’ RAN Intelligent Controller (RIC), which will be offered for free.
In the near term, Symphony says it sees no need to use another RIC besides Juniper’s and no need to use another SoC besides Intel’s.
To the impartial observer, the idea of Symphony forming de facto exclusive relationships with Intel and Juniper might seem to contradict the Open vRAN vision of choosing a different vendor to supply each element in a disaggregated RAN. And the idea of using integrated accelerators to optimize processors for RAN workloads might seem to contradict the notion of using general-purpose silicon for vRAN.
But these moves – like the decision to give away Juniper’s RIC – are meant to aid an industry segment that still needs a push to get going. A pure-play disruptor like Symphony is highly motivated to get this space rolling, and its odds of success are greater with the help of heavyweights like Intel and Juniper.