Remember when everyone was obsessed with dedicated silicon to speed up the performance of commercial off the shelf (COTS) servers dealing with 5G cloud radio access networks (C-RAN) workloads, rather than the joys of generative artificial intelligence (AI)?
Well, AvidThink analyst Roy Chua does. So, Silverlinings asked him for a quick update on the 5G acceleration game.
Intel still in the lead
“Intel is still the lead provider for L1 acceleration,” Chua said. Many mobile network operators have already deployed virtualized radio access network (vRAN) networks using Intel Xeon chips, from Verizon with thousands of vRAN sites already deployed, thanks to Samsung. To Rakuten deploying its large Open RAN deployment with Intel chips.
Most mobile network operators (MNOs) haven’t moved on to Intel’s 4th generation Xeon chips with an integrated vRAN boost yet. But Intel still leads the pack in the L1 game with its “lookaside” x86 acceleration, which was based on field programmable gate array (FPGA) cards and its FlexRAN software.
Marvell, Nvidia and Qualcomm
Chua said that Marvell offers everything from Open RAN to legacy RAN. He notes that the chip designer supports Nokia with its “inline” L1 acceleration, noting that “they have feet in every boat.”
Chua notes that Qualcomm is also looking to get into the market late this year or early next, starting with operator trails.
Nvidia, he notes, is also trying to provide its GPU processors to deal with everything from AI to L1 RAN acceleration.
Like others, Chua said that the march to the 5G RAN market appears to be cooling somewhat.
“Now that you see the 5G build-up slowing down a little bit, and the open RAN rollout slow down a little bit, I think that Intel, Marvell, Qualcomm, and to a certain extent, Nvidia, will need to be cautious about forward investment,” he said.
Want to learn more about moving core networks to the cloud? Don’t miss our upcoming Telco Core Strategies Summit on Sept. 18. Learn more and register here.