Home News High-performance module for physical AI applications in edge environments...

High-performance module for physical AI applications in rugged edge environments

High-performance module for physical AI applications in rugged edge environments

congatec today introduced its new COM-HPC Client Size C module featuring AMD Ryzen™ AI Embedded X100 Series processors. The out-of-the-box aReady.COM conga-HPC/cRX1 module is specifically designed for high-performance physical AI applications. With up to 16 AMD “Zen 5” CPU cores, up to 40 compute units on the integrated AMD Radeon RDNA 3.5 GPU, and a dedicated NPU, the conga-HPC/cRX1 sets a new performance benchmark for COM-HPC Client modules. Its high-performance integrated AI and FP32 compute capabilities eliminate the need for discrete AI accelerator cards in many target applications. Combining exceptional computational performance, wide I/O bandwidth, and support for industrial temperatures ranging from -40°C to +85°C, the module is ideal for physical AI applications that integrate perception, intelligence, and actuation to execute complex tasks in dynamic operating environments. Typical application areas include robotics, autonomous commercial vehicles for smart agriculture, logistics and forestry, as well as electromedicine and industrial automation.

The COM-HPC module, part of the AMD Kria™ AI SOM module family, leverages the new AMD Ryzen AI Embedded X100 series, with up to 16 "Zen 5" CPU cores, for time-critical sequential workloads such as real-time control, sensor fusion, planning, and decision-making. The integrated GPU delivers up to 59 TOPS of dense AI inference performance in INT8 and up to 29,7 TFLOPS of compute performance in FP32, accelerating highly parallel workloads such as object perception and detection in collaborative robots and autonomous vehicles. It also enables local execution of large language models (LLMs). The dedicated NPU further provides up to 50 TOPS of AI performance for always-on AI workloads such as object detection and recognition, speech recognition, and image processing. Based on scalable COM (Computer-on-Modules) technology, congatec solutions enable the creation of robust, modular embedded platforms with long-term availability that combine low system complexity with maximum computing performance.

“Physical AI applications are placing ever-increasing demands on embedded systems,” says Florian Drittenthaler, Productline Manager at congatec. “Our conga-HPC/cRX1, equipped with AMD Ryzen AI Embedded X100 Series processors, delivers the highest performance currently available in a COM-HPC Client module. This allows original equipment manufacturers (OEMs) to forgo discrete AI accelerator cards in many applications, enabling them to develop more compact, energy-efficient, reliable, and cost-effective edge embedded systems with a lower total cost of ownership.”

“The next wave of embedded systems innovation will be characterized by delivering more intelligence in smaller, more energy-efficient designs,” said Sumit Shah, Head of Product Management and Marketing, Adaptive and Embedded Computing Group, AMD. “AMD Ryzen AI Embedded X100 Series processors integrate CPU, GPU, and NPU capabilities into a single platform, reducing system complexity while delivering the deterministic performance and efficiency needed for AI in the real-world edge environment.”

Features in detail

The conga-HPC/cRX1 is available in the COM-HPC Client Size C form factor, measuring 120 mm × 160 mm, and supports industrial temperatures from -40 °C to +85 °C, making it suitable for deployment in harsh operating environments. With a base TDP of 55 W, the module can be configured across a wide TDP range, from 45 W for optimized energy efficiency to 120 W for maximum performance. The AMD Ryzen AI Embedded X100-series processors feature up to 16 AMD “Zen 5” CPU cores operating at clock speeds up to 5,1 GHz. The integrated AMD Radeon RDNA 3.5 iGPU supports up to four independent 4K displays, while the dedicated AMD XDNA 2 NPU delivers up to 50 TOPS of AI performance for highly efficient parallel AI computing.

Memory-intensive applications benefit from up to 128 GB of unified LPDDR5X-8533 memory, enabling efficient data exchange between compute engines. This reduces latency while simplifying system design. Furthermore, the memory is soldered to increase the module's resistance to shock and vibration. The capacity and high bandwidth of the unified memory accelerate general-purpose GPU (GPGPU) workloads, which benefit from high data throughput between memory and compute engines. Additionally, up to 512 GB of optional integrated NVMe storage provides fast data access and reduced load times. For connecting expansion devices, the module offers up to 24 PCIe Gen4 lanes, allowing easy integration of numerous low-bandwidth devices, such as industrial Ethernet interfaces, fieldbus adapters, or wireless communication modules. Additional interfaces include two 2,5 GbE ports, up to four USB 3.2 Gen2 ports, up to eight USB 2.0 ports, up to two SATA 6 Gb/s ports, two I²C ports, two UART ports, twelve GPIO pins, one SMBus port, and one SPI port. The module has been developed in accordance with IEC 62443-4-1. This helps original equipment manufacturers (OEMs) meet cybersecurity requirements and simplifies their preparations for the EU Cyber ​​Resilience Act, which will come into effect in December 2027. A Trusted Platform Module (TPM 2.0) provides additional built-in security. To accelerate the development and deployment of AI applications, the comprehensive AMD ROCm open-source software stack provides developers with a powerful platform for AI workloads and high-performance computing.

Operating systems compatible with the conga-HPC/cRX1 include Microsoft Windows 11, Windows 11 IoT Enterprise, and Linux. As an application-ready aReady.COM module, the conga-HPC/cRX1 is also available pre-configured with the licensed operating systems ctrlX OS, Ubuntu Pro, or KontronOS. The aReady.VT option, which features the integrated conga-zones hypervisor and virtualization technologies, allows developers to consolidate multiple workloads—such as real-time control, human-machine interface (HMI), AI, and IoT gateway functions—into a single module. For IIoT connectivity, congatec offers the aReady.IOT software components, including conga-connect, to enable data sharing, remote maintenance, and management of the module, motherboards, and peripherals as needed. To further simplify application development, congatec offers a comprehensive support ecosystem that includes evaluation and application motherboards, custom cooling solutions, documentation, design integration services, and high-speed signal integrity measurements.

With aReady.YOURS, original equipment manufacturers (OEMs) can also take advantage of congatec's extensive hardware and software customization services to obtain "turnkey" embedded computing platforms that include advanced cooling solutions.

Summary of the new variants of the conga-HPC/cRX1 processor:

Córdoba Cores / Threads Graphics execution units Clock frequency (max. boost) Base TDP (configurable) Operating temperature
AMD Ryzen AI Embedded X199 16 / 32 40 5.1 GHz 55W (45-120W) 0 to +60°C
AMD Ryzen AI Embedded X188 12 / 24 32 5.0 GHz 55W (45-120W) 0 to +60°C
AMD Ryzen AI Embedded X168 8 / 16 32 5.0 GHz 55W (45-120W) 0 to +60°C
AMD Ryzen AI Embedded X199i 16 / 32 40 5.1 GHz 55W (45-120W) -40 to +85°C
AMD Ryzen AI Embedded X188i 12 / 24 32 5.0 GHz 55W (45-120W) -40 to +85°C
AMD Ryzen AI Embedded X168i 8 / 16 32 5.0 GHz 55W (45-120W) -40 to +85°C

For more information on the new conga-HPC/cRX1 computing modules, please visit: https://www.congatec.com/en/products/com-hpc/conga-hpccrx1/