The AMD Ryzen™ AI Embedded X100 Series processors define a new performance class for scalable embedded and edge computing. They combine up to 16 AMD “Zen 5” CPU cores, an integrated Radeon™ RDNA™ 3.5 iGPU, and a dedicated XDNA™ 2 NPU delivering up to 50 TOPS of AI acceleration within a System-on-Chip (SoC) with high compute density. Highly integrated AI and FP32 compute performance eliminate the need for discrete AI accelerator cards in many target applications.

Scalable Performance for Physical AI at the Edge

The AMD Ryzen™ AI Embedded X100 Series processors are part of the AMD Kria™ AI SOM portfolio and have been specifically designed to address the growing performance requirements of modern physical AI applications. They combine powerful CPU, GPU, and NPU resources within a high-performance SoC, enabling deterministic AI performance directly at the edge. As a result, there is no need for discrete graphics cards, additional AI hardware accelerators, or cloud processing.

The new AMD Ryzen™ AI Embedded X100 Series features up to 16 AMD “Zen 5” CPU cores alongside an integrated AMD Radeon™ RDNA™ 3.5 GPU that delivers up to 59 TOPS of AI inference performance and up to 29.7 TFLOPS of FP32 compute performance. With a dedicated AMD XDNA™ 2 NPU providing up to 50 TOPS of additional AI performance, these SoCs provide a highly scalable platform for a broad range of AI workloads. They are particularly suited for demanding physical AI applications in robotics, industrial automation, medical technology, autonomous commercial vehicles, and smart infrastructure.

Built on AMD's advanced “Zen 5” architecture using 4 nm process technology, the SoCs deliver exceptional CPU compute performance for time-critical and sequential workloads such as real-time control, sensor fusion, and decision-making. The integrated RDNA™ 3.5 GPU is ideally suited for highly parallel workloads such as image pre-processing, perception and object recognition in collaborative robots and autonomous vehicles, simultaneous localization and mapping (SLAM), and the local execution of large language models (LLMs). The XDNA™ 2 NPU complements the architecture with a highly efficient AI accelerator optimized for always-on AI applications such as object detection, speech processing, image classification, and anomaly detection.

A key advantage for generative AI applications is the unified memory architecture. The CPU and GPU share the same high-speed memory pool, allowing scaled memory allocation to the GPU based on the application. With up to 128 GB of LPDDR5x memory, as much as 96 GB can be made available for GPU-driven workloads, enabling large LLMs to run locally without requiring discrete graphics cards equipped with dedicated memory.

For Scalable High-Performance Physical AI Systems

With a widely configurable TDP range from 45 to 120 W, the new SoCs enable flexible embedded compute optimization for a broad variety of performance and cooling concepts – from edge applications optimized for low power consumption to compute-intensive physical AI platforms. Developers additionally benefit from the scalable COM architecture, which allows underlying hardware design to be preserved across related modules with different performance levels.

The high level of performance provided by the AMD Ryzen™ AI Embedded X100 Series processors eliminates the need for discrete graphics or dedicated AI accelerator cards in many applications. This simultaneously reduces system complexity, power consumption, cooling requirements, and bill of materials (BOM) costs. At the same time, a design with fewer critical components improves overall application reliability and robustness.