In an open RAN progress report published last November, Ericsson acknowledged its dependency on the big US chipmaker. In a table listing the available suppliers in different parts of the value chain, Intel stood out as the sole "commercial-grade" option for central processing units (CPUs), accelerators, and network interface cards (NICs).
According to that same report, Ericsson had reached only the "active engagement" stage with AMD and its Xilinx subsidiary. No other chipmaker was named.
While this might be okay normally, Chipzilla’s recent woes worry the Son of Eric executives—particularly if Intel is sold to the highest bidder. Ericsson was itself linked to a possible purchase of RAN assets for this reason.
Ericsson and its rivals continue to invest in the purpose-built chips used in most of today's 5G networks, a market also served by Intel, Marvell Technology, and other merchant silicon vendors. However, some Arm licensees offer a design alternative to Intel's x86 architecture and offer an opportunity in virtual RAN. Among them is Nvidia. Its Grace Hopper and Grace Blackwell offer package Arm-based CPUs with graphical processing units (GPUs) for what Nvidia calls an AI RAN.
Ericsson's response is to push for as much separation of software from hardware as possible.
Ericsson's mobile networks business group boss Fredrik Jejdling said he didn’t know how it would pan out, but harmonising software has solved some of the problems.
The company has stood by what Jejdling describes as a "lookaside" approach to virtual RAN. That means nearly all the RAN software is handled by a general-purpose CPU, including most of the network functions in "Layer 1," the software slice which needs the most IT resources.
"Our Layer 1 can run on different types of underlying hardware, and that is exactly what we offer actual customers: true cloud RAN through virtualization," said Jejdling.
The alternative, historically known as "inline," shifts the whole of Layer 1 to a separate accelerator, where there is the tighter linkage between hardware and software.
Nokia, the former rubber boot maker and Ericsson’s biggest rival, prefers it and runs its Layer 1 software on a chip co-developed with Marvell. The NIC hosting this chip can be connected to any server with a standard PCIe interface. While the higher-layer software looks as independent of hardware as Ericsson's, switching to another Layer 1 silicon vendor would necessitate a rewrite of code.
But full software independence is still tricky. One problem is using accelerators to cope with forward error correction (FEC), an incredibly demanding part of the Layer 1 software stack.
Ericsson relies on Intel's FEC accelerator in virtual RAN deployments with the chipmaker. The accelerator is not even if the rest of the Layer 1 code is reusable with an alternative CPU. In Sapphire Rapids EE, Intel's latest generation, it integrates with the CPU, rather than on a separate NIC. According to Intel's roadmap, the forthcoming CPU products, beamforming, and another Layer 1 function will also move to the accelerator in Granite Rapids and Diamond Rapids.
Jejdling concedes that some Layer 1 alterations would be necessary if the Sons of Eric moved from Intel to another chipmaker. Still, he insists the "lion's share" of the stack is "common" across platforms. "In Barcelona two or three years back, we took a cloud RAN stack at the time, built on Intel, and we ported it over to AMD with quite little change even in Layer 1," he said.
AMD executives have mocked Intel's recent accelerator moves, which they think are too limiting and difficult to scale.
AMD uses the same x86 architecture as Intel. Everyone else falls into the Arm camp, and nobody except Ericsson has talked publicly about applying the look-aside approach to an Arm processor. Even Ericsson admitted about 18 months ago that Arm had seemed to lack the necessary power for Layer 1.
At the time, it was hoped that an architectural update called SVE2 would do for Arm processors what a technology known as AVX-512 has done for Intel. Yet Ericsson still appears not to be in "active engagement" with any Arm licensees over virtual RAN so that probably went nowhere.
From Oracle-backed Ampere Computing to Qualcomm, those Arm licensees have all backed the inline approach instead. Some, like Marvell and Qualcomm, have produced Layer 1 accelerators that feature Arm cores, digital signal processors, and other custom hardware combinations. Others like Ampere and AWS are developing Arm-based CPUs for higher-layer functions.
Nvidia is the world's most highly valued Arm licensee. It also prefers inline architecture and wants to see Layer 1 plus certain higher-layer functions handled by a Hopper or Blackwell GPU. It has written its own RAN software, Aerial, which supports Layer 1 for a trial deployment with SoftBank in Japan. Fujitsu is responsible for providing higher-layer software and radio units.
Analysts think purpose-built RAN will remain the default choice for many years. Dell'Oro predicted it would account for more than three-quarters of the entire global market by 2029.
