Atoosa Hatefi, head of radio innovation at Orange Group, is convinced that O-RAN-compliant radio units (O-RUs) can at least match the energy efficiencies of traditional RUs.

She observes, too, large energy-efficiency strides being made within the cloud RAN ecosystem to support virtualized baseband unit (BBU) functions – distributed unit (DU) and centralized unit (CU) – running on commercial-off-the-shelf (COTS) hardware combined with accelerators. (Cloud RAN, for Hatefi, can be viewed as a "subset" of open RAN.)

"We're confident that open RAN's green credentials can at least be on a par [with] and even better than traditional RAN, because there are so many levers we can pull to allow open RAN to become better," she told Light Reading.

Among these levers, asserts Hatefi, are much faster-paced development cycles for cloud infra chipsets versus purpose-built hardware – normally a two-year cycle, she says, rather than typically double that found in traditional RAN deployments. "Cloud RAN," anticipates Hatefi, "may overtake traditional BBUs."

Another open RAN green lever, she says, is pooling gains courtesy of a new and more centralized server architecture, coupled with "dynamic switch on/off of hardware resources."

Moreover, argues Hatefi, radio intelligent controller (RIC) apps will help further improve cloud RAN energy efficiency. "Although legacy equipment can also take advantage of RIC apps, open interfaces mean those apps will be more natively supported and could be therefore more widely adopted within the open RAN ecosystem than with traditional RAN."

Specifications for three energy-saving RIC use cases, expects Hatefi, will be finalized by the O-RAN Alliance before the year is out: carrier and cell switch off/on for near-real-time RIC (already completed for non-real-time RIC); RF channel reconfiguration; and advanced sleep modes.

"At some point of time, with all these levers to pull – and which are not available with traditional RAN – energy-efficiency performance will exceed that of closed RAN," asserts Hatefi.

What about that power-pumped radio unit…

The RU, thanks to the energy-hungry power amplifier, accounts for around 80% of total power consumption at a radio site. One school of thought is that O-RU newcomers will always be handicapped by much smaller R&D budgets than the "big guys" and in all likelihood will never be able to compete meaningfully here. It's not a view that Hatefi subscribes to.

"There's no particular reason for an open radio unit to be less power efficient than a closed RU, as it uses more or less the same type of architecture," she says. "While there is openness of the interface, the design and manufacture of the O-RU uses similar types of processes and functionalities found in traditional RAN. Even today we are seeing O-RAN compliant RUs on a par with closed RUs in terms of energy efficiency."

Hatefi does acknowledge that the more energy-efficient O-RAN-compliant RUs are currently coming from traditional suppliers, but she sees that gap as only temporary and closing. Once some of the new players win bigger contracts and their volumes increase, thinks Hatefi, a virtuous cycle will emerge allowing them to further increase energy-efficiency optimization.

She already sees "lots of progress" among some new RU challengers in terms of greater power‑amplifier efficiencies and smarter algorithms. "And open RAN suppliers," Hatefi pointedly adds, "have access to the same third-party providers as traditional RAN vendors to develop their products." It's Hatefi's contention that the RU product in open RAN and traditional RAN "may be exactly the same with the same energy efficiency."

James Crawshaw, a principal analyst at Omdia, a Light Reading sister company, remains unconvinced that open RAN will have a steeper green trajectory than traditional RAN, mainly because the power amplifier – "the key energy hog in the entire mobile system" – is not going anywhere anytime soon.

"Open RAN is about disaggregating elements and introducing open interfaces so it's easier to mix and match suppliers, but none of that helps make the power amplifier more energy efficient," he told Light Reading.

…and hardware performance lag?

As with O-RU energy performance, Hatefi acknowledges that there has been an energy-efficiency performance gap between virtualized RAN software running in a cloud RAN environment and purpose-built RAN. "But as with open RAN," she caveats, "there's been a lot of progress in the cloud RAN ecosystem. For low-capacity scenarios in rural deployments, the power consumption performance between traditional RAN and cloud RAN is already getting similar."

Hatefi is confident, courtesy of a new generation of chipsets, that cloud RAN will start matching – from 2025 onwards – the energy-efficiency performance of traditional RANs in high-capacity urban scenarios: namely, massive MIMO running on generic hardware combined with accelerators.

"That's because there's been lots of progress, not only with chipsets and accelerators, but also in the dimensioning part," she says. "The newest generation of chipsets with higher core counts should be able to sustain even the highest-capacity scenarios with a mix of FDD bands and TDD bands in massive MIMO in a single server."

Observing chipset developments from various suppliers, and noting RAN advances made by GPU (graphics processing unit) powerhouse Nvidia, Hatefi expects the performance crossover point between dedicated and generic hardware will happen sometime during 2025 or early 2026.

"The plan from next year is to be able to consider only one server per site for high-capacity scenarios," continues Hatefi. "This will be the main lever to optimize the energy consumption of cloud RAN to be on a par with traditional RAN."

As with the O-RU, Omdia's Crawshaw doubts that open RAN proponents will be able to make much green hay in generic hardware, unless there are considerable improvements in CPU (central processing unit) energy performance, assisted perhaps by some ASICs (application-specific integrated circuits) on the side for network acceleration.

"Open RAN is largely about using x86 general-purpose CPUs instead of specialized chips and I think it'll be very hard for CPU-based systems to be more energy efficient than ASIC-based systems," says Crawshaw. "I'm by no means dismissing the open RAN movement. It's just that I don't think better energy efficiency is a strong argument in favor."