High-Performance Embedded Compute Modules with Scalable AI Acceleration

Congatec has introduced a new portfolio of Computer-on-Modules (COMs) built around Intel Core Ultra Series 3 processors, offering up to 180 TOPS of heterogeneous computing performance designed to accelerate embedded AI workloads without requiring discrete accelerator hardware — a key consideration for size, weight, and power-constrained industrial and edge systems. These modules address technical requirements in sectors such as industrial automation, autonomous robotics, smart infrastructure, healthcare imaging, and point-of-sale systems.

Integrated Heterogeneous Computing Architecture
The Intel Core Ultra Series 3 processors at the heart of the new modules implement a heterogeneous architecture comprising up to 4 performance (P) cores, up to 8 efficiency (E) cores, and up to 4 low-power E-cores, coupled with an integrated neural processing unit (NPU5) and up to 12 Xe3 GPU cores. This combination delivers up to approximately 10 TOPS from general-purpose CPU execution, up to 50 TOPS from the NPU for low-power inference, and up to ~120 TOPS from the GPU for highly parallel workloads, totalling up to 180 TOPS of AI acceleration potential. The integrated compute engines aim to support local AI tasks such as natural language processing (NLP), large language model (LLM) execution, image classification, sensor fusion, and simultaneous localization and mapping (SLAM) without requiring discrete AI accelerators.

Intel’s 18A semiconductor process underpins this processor generation, employing RibbonFET transistors and PowerVia backside power architecture to increase transistor density and performance-per-watt relative to previous technology nodes. The heterogeneous compute architecture, alongside support for standards-based frameworks such as Intel^® Xe Matrix Extensions (XMX) and OpenVINO, enables optimisation of local machine learning models on embedded systems without off-board processing.

Modular Form Factors and Interfaces
The new congatec COM portfolio spans multiple standard form factors to support both new and legacy system designs. Smaller, power-optimised modules in COM Express Mini and COM-HPC Mini form factors serve space-constrained applications, while COM Express Compact modules — including ruggedised variants with screw-locked LPCAMM2 memory — provide a cost-sensitive upgrade path for existing deployments. For high I/O throughput and extensibility, the COM-HPC Client modules deliver PCIe Gen5 and USB4 interfaces critical for data-intensive machine vision, sensor integration, and high-bandwidth communications.

Each module supports up to 96 GB of LPDDR5X memory (soldered) or socketed LPCAMM2 memory with optional in-band Error Correction Code (ECC), and integrated graphics supporting up to three independent 6K displays. Operating systems supported include Microsoft Windows 11 and Windows 11 IoT Enterprise, Ubuntu Pro, Yocto, and vendor-specific real-time variants such as ctrlX OS and KontronOS.

Software and Ecosystem Support
To accelerate development, congatec bundles its modules with software building blocks and tools. The aReady.COM option offers pre-configured software stacks, while aReady.VT delivers a hypervisor-on-module for consolidating workloads such as real-time control, human-machine interface (HMI), AI inference, and IoT gateway functions on a single module. The aReady.IOT components enable industrial IoT connectivity and device management. An ecosystem of evaluation and production carrier boards, cooling solutions, documentation, and design-in services further supports system integration and time-to-market.

Application Context and Technical Relevance
The introduction of Intel Core Ultra Series 3-based modules responds to rising demand for embedded AI performance at the edge, where constraints on power, space, and thermal dissipation often preclude the use of discrete accelerators. By integrating significant AI acceleration alongside general-purpose processing and high-bandwidth interfaces, these modules aim to simplify system architecture in robotics, industrial automation, smart facility management, advanced imaging systems, and intelligent retail terminals, while preserving long lifecycle support and reliability critical in embedded and industrial applications.

www.congatec.com