Modules from German firm Congatec are based on AMD processors and are designed for edge AI applications in transportation, medical, smart city infrastructure, gaming, point-of-sale, robotics and industrial automation.

The Conga-TCRP1 Com Express 3.1 Type 6 compact modules are based on the AMD Ryzen AI Embedded P100 processors with four or six cores, and support temperatures from -40 to +85˚C.

They accelerate embedded computing applications with up to 59Tops of combined AI inference performance, with up to 50Tops provided by the XDNA2 NPU, and the remaining performance delivered by up to six AMD Zen5 CPU cores and the RDNA3.5 GPU. Those who want to balance their application between efficient CPU, GPU and NPU performance, configurable thermal design power (TDP) from 15 to 54W, and size, weight, power and cost requirements could benefit from the scalability of the modules.

The minimum configurable TDP of 15W simplifies the development of passively cooled, completely enclosed designs. This makes them suitable for use in rugged handheld devices, hygienic medical PCs and mission-critical devices in harsh environments.

“With the increasing demand for rugged edge applications with or without AI, there is also a growing need for entry-level Com variants with four and six cores for power- and cost-optimised designs,” said Florian Drittenthaler, product line manager at Congatec. “The Conga-TCRP1 expands Congatec’s comprehensive portfolio for precisely these application areas, targeting cost-sensitive designs that require optimal performance per watt.”

The computing cores are based on the AMD Zen5 architecture with performance-oriented Zen5 cores combined with Zen5c cores designed for energy efficiency. This ensures low power consumption while delivering high overall efficiency and a single-thread performance of up to 4.5GHz. An advantage for application developers is that the Zen5 and Zen5c cores are based on the same architecture. This ensures consistent execution timing in deterministic applications and optimises real-time performance.

In addition to the Zen5 and 5c cores and the Radeon RDNA 3.5 GPU, which supports up to four independent display connections with immersive 4k graphics, the integrated XDNA2 NPU provides up to 50Tops of AI performance. This lets it run real-time processing of smaller large language models (LLMs) locally, without cloud connectivity or discrete accelerators, and in a cost- and energy-efficient manner.

At the same time, memory-intensive applications benefit from up to 96Gbyte of DDR5-5600 RAM with optional error correction code (ECC) for mission-critical applications. Up to eight fully configurable PCIe Gen4 lanes and PEG x4 Gen4 are available for fast data transfer and the connection of low-lane peripherals such as industrial Ethernet, Fieldbus adapters or radio modules.

In addition, 2.5GbE ensures fast networking, and four USB 3.2 Gen2 and four USB 2.0 provide connections for additional devices. For data storage, the modules offer up to 512Mbyte of onboard NVMe SSD or two Sata 6Gbit/s for external media. One I²C bus, GPSPI, two uarts, eight GPIOs, one SMBus and one LPC round out the feature set.

Supported operating systems include Microsoft Windows 11, Windows 11 IoT Enterprise, Linux, CtrlX OS, Ubuntu Pro and Kontron OS. The modules can be preconfigured with licensed CtrlX OS, Ubuntu Pro and Kontron OS.

For IoT connectivity, Congatec offers aReady.IoT software building blocks. These enable data exchange, remote maintenance and management of the module, carrier board and peripherals, as well as cloud connectivity, if required. To simplify application development, Congatec (www.congatec.com) offers an accompanying ecosystem that includes evaluation and application-ready carrier boards, customised cooling, documentation, design-in services, and high-speed signal integrity measurements.

For more on the modules, visit www.congatec.com/en/products/com-express-type-6/conga-tcrp1/.