Accelerated Edge Inference on Intel Silicon
vRAN workloads, FlexRAN baseband processing, and AI model deployment at the far edge — hardware-optimised for Intel Xeon and Core Ultra with maximum performance per watt.
Talk to an ExpertThe Full Intel Edge AI Stack, Optimised by OranSense
Intel's Xeon Scalable and Core Ultra processors bring hardware AI acceleration to the network edge — AVX-512, AMX matrix extensions, and integrated GPU compute enable vRAN and AI inference workloads to run on the same silicon without dedicated accelerator cards.
OranSense optimises the full software stack for Intel hardware — from FlexRAN baseband libraries and DPDK data plane acceleration to OpenVINO model inference and OPNFV virtualisation. The result is a software-defined RAN and AI inference platform that delivers carrier-grade performance on commercial off-the-shelf Intel servers.
Key Capabilities
FlexRAN Baseband Processing
Intel FlexRAN reference software optimised for Xeon Scalable. Run 5G NR Layer 1 processing — LDPC encode/decode, PUSCH/PDSCH — on CPU with hardware acceleration via Intel ACC100/ACC200 FEC cards.
OpenVINO AI Inference
Deploy AI models optimised for Intel hardware using OpenVINO toolkit. Achieve 3x inference throughput versus generic frameworks on the same silicon with INT8 quantisation and hardware-specific graph optimisation.
DPDK Data Plane Acceleration
Intel DPDK enables kernel-bypass packet processing for fronthaul and midhaul interfaces. Achieve line-rate packet forwarding on Xeon with deterministic sub-microsecond latency.
vRAN on COTS Hardware
Run a complete virtualised RAN stack — CU, DU, and RU interface — on Intel Xeon servers. Eliminate proprietary RAN hardware and reduce TCO by up to 50% versus traditional base stations.
Power-Efficient Far Edge Deployment
Intel Core Ultra enables AI inference and lightweight RAN workloads in thermally constrained far-edge locations — street cabinets, lamp posts, and enterprise premises — at under 15W TDP.
Intel Network Platform Integration
Native integration with Intel Network Platform — E810 SmartNIC, QuickAssist Technology for crypto/compression, and Time Coordinated Computing for precision timing in fronthaul networks.
NVIDIA & Intel Silicon, Unified by OranSense
From far-edge Jetson modules to centralised DGX AI factories — OranSense optimises the full software stack across every hardware tier.
Jetson AGX Orin
Purpose-built for autonomous edge AI — Jetson AGX Orin delivers 275 TOPS of AI compute in a compact, thermally efficient module. OranSense deploys vRAN inference, xApp execution, and real-time RF analytics directly on Jetson at the far edge: street cabinets, drones, and enterprise premises.
DGX H100
NVIDIA DGX H100 anchors the OranSense AI Factory — powering large-scale model training, digital twin simulation, and high-throughput inference for centralised RAN intelligence. Eight H100 GPUs with NVLink 4.0 fabric deliver the compute density required for real-time network-wide AI at scale.
Xeon Scalable (Sapphire Rapids)
Intel Xeon Scalable with AMX matrix extensions and integrated ACC200 FEC acceleration runs the full 5G NR Layer 1 stack — LDPC encode/decode, PUSCH/PDSCH — without a dedicated GPU. OranSense optimises the FlexRAN and DPDK stack for carrier-grade DU deployments on COTS servers.
Core Ultra (Meteor Lake)
Intel Core Ultra brings an integrated NPU and GPU to thermally constrained far-edge locations. OranSense leverages OpenVINO and the Intel AI Engine for lightweight xApp inference and RF anomaly detection at enterprise premises and outdoor micro-cells — all within a 15W power budget.