Intel FPGA Accelerator Targets 5G Rollouts
Intel unveiled an FPGA-based programmable acceleration card aimed at 5G wireless service providers handling growing volumes of virtualized workloads.
The N3000 acceleration card was rolled out during the first day of the Mobile World Congress in Barcelona, Spain. The chip maker is betting the ongoing 5G rollout and a shift toward network functions virtualization (NFV) will place a premium on a combination of low-latency FPGAs and the greater bandwidth the new wireless standard will provide for connecting edge devices and the resulting flood of new data traffic.
Intel said this week its accelerator combines low-latency FPGAs with “enhanced mobile broadband” that together deliver a bigger data pipeline for machine-to-machine connections. Those links are expected to contribute to a huge spike in data traffic, creating what Intel dubs a “5G provider crisis.”
Hence, the accelerator card is aimed at 5G service providers scrambling to deliver “core” and virtualized radio access frameworks along with emerging NFV capabilities. The company said its FPGA approach would help speed virtualized workloads ranging from core network applications to 5G radio access networks.
An Intel partner, Affirmed Networks, said it was able to cut CPU utilization in half while handling a data traffic loads based on its cloud-native, application container setup. “5G is a transformative technology, and it requires advanced network virtualization infrastructure coupled with an agile software architecture,” said Ron Parker, chief architect at Affirmed Networks.
The Intel partner said it developed a 5G core network platform based on Intel’s FPGA accelerator card that yielded 200 Gbps per server performance. FPGAs were used for “smart” load balancing and CPU cache optimizations.
Another Intel partner, Japanese mobile network operator Rakuten, integrated the FPGA accelerator card into its planned cloud-native mobile network.
In targeting high-bandwidth applications, the N3000 is designed to boost data plane performance. The result is said to be an acceleration of network traffic up to 100 Gbps with support for in-memory processing of high-end applications.
Last April, Intel announced broad adoption of its programmable acceleration card running its Arria 10 GX FPGA in Dell-EMC and Fujitsu servers.