Intel Adds VRAN Acceleration To Its Latest 4th Gen Xeon Scalable SoCs

At this week’s MWC, Intel announced new, communications-centric members of its latest Xeon processor generation, the 4^th Gen Xeon Scalable Processors with Intel vRAN Boost. These System on Chips (SoCs) have been designed with extensions specifically tailored to accelerate packet- and signal-processing workloads in virtual radio access networks (vRANs). These processors are also designed to accelerate other workloads commonly run in vRAN environments such as load balancing, AI and Machine Learning (AI/ML), and dynamic power management. The company claims that these accelerated Xeon processors can deliver equal or superior performance-per-watt when compared with “the best” Layer 1 (physical layer) SOC accelerator cards currently on the market.

These 4^th Gen Xeon SoCs are designed for use by Communications Service Providers (CoSPs) that are building out their 5G networks using vRANs. Intel claims that 99% of today’s vRAN deployments already use Intel Xeon processors running the company’s FlexRAN software. The 4^th Gen Xeon SoCs will run this software unchanged and can deliver power and performance benefits solely from the generation-on-generation improvement, but the new Xeon Scalable processors incorporate specialized instruction extensions called Advanced Vector Extensions (AVX) for vRAN that allow the 4^th Gen Xeon processors to enable twice the vRAN traffic capacity within the same power envelope when compared to the company’s 3^rd Gen Xeon Scalable devices. Further, Intel claims that these extensions eliminate the need for companion vRAN accelerator cards, which further reduces power consumption by as much as 20% and reduces vRAN Total Cost of Ownership (TCO).

Although these accelerated 4^th Gen Xeon Scalable SoCs can directly run existing FlexRAN code, realizing the full benefit of the extensions requires a new version of the FlexRAN software. Intel says it has made such modifications to its FlexRAN software to support a wide range of 5G scenarios including sub-6GHz, mmWave, small-cell, and private wireless network deployments. The company is also supporting independent software vendors (ISVs) to allow them to modify their software to take advantage of the new vRAN extensions through a suite of updated software development tools, libraries, and software kits. One necessary vRAN feature not implemented in these new 4^th Gen Xeon Scalable SoCs is support for the IEEE-1588 Precision Time Protocol (PTP). Instead, Intel has chosen to rely on its E810 series of Ethernet adapters for the PTP support.

Intel’s decision to add vRAN-specific extensions to this group of 4^th Gen Xeon Scalable SoCs reflects the company’s strategy of making it possible to enable all servers in a CoSP’s data center to implement vRANs. If all processors in the data center incorporate these extensions, then the CoSP can scale its vRAN footprint up or down in the data center according to the subscriber traffic load. This approach eliminates the need to augment a subset of the data center’s servers with accelerator cards just to accommodate the processing needs of the vRAN workload, which should simplify data center maintenance and upgrade strategies.

Several of Intel’s system customers including Advantech, Canonical, Ericsson, Quanta Cloud Technology, and Supermicro are already supporting or are signed up to support these 4^th Gen Xeon Scalable processors with vRAN extensions. They’ve likely found the hardware conversion to be particularly easy because the processors with vRAN extensions are socket-compatible with other 4^th Gen Xeon Scalable processors.

Adding these vRAN extensions to the 4^th Gen Xeon Scalable processors is consistent with Intel’s current philosophy of continually upgrading the Xeon architecture with accelerators to meet the evolving needs of its customers. For additional perspectives on that subject, see my colleague Jim McGregor’s recent article titled “Sapphire Rapids Is A Critical Turning Point For Intel” on Forbes.com.