AMD is expanding its Ryzen AI Embedded P100 Series with new processors offering up to 12 Zen 5 cores, 80 system TOPS, and up to 8x higher GPU compute. The new chips target industrial automation, mobile robotics, and medical imaging. In addition, they now include AMD ROCm software support for open-source AI development at the edge.
Compared to the previously announced P100 chips, the new additions feature up to 2x more CPU cores, up to 8x higher GPU compute, and an estimated 36% higher system TOPS. This is all within the same compact package. Against the older Ryzen Embedded 8000 Series, the performance gains are even more substantial: up to 39% higher multithreaded performance and 2.1x more total system TOPS.
The chips support industrial temperature ranges from -40°C to 105°C, 24/7 operation, and 10-year product life cycles. AMD also says the new P100 processors can support almost twice the number of virtual machines and run larger LLMs, such as Llama 3.2-Vision 11B, compared to the existing P100 lineup. What constitutes ‘larger’ is obviously relative, these are not datacenter GPUs, but for their industrial use cases models of a few billion parameters are often easily sufficient.
Three use cases, one chip
The expanded P100 processors integrate eight to 12 Zen 5 cores, AMD RDNA 3.5 graphics, and an XDNA 2-based NPU on a single chip. AMD focuses on three main application areas. For industrial PCs, the chip consolidates PLCs, machine vision, and HMI (human-machine interface) into one platform. They are geared for running anomaly detection models, benchmarks often using PaDiM and Llama 3.2-Vision. For mobile robots, unified CPU-GPU memory reduces latency, while the always-on NPU handles object detection with models like YOLOv12. In medical imaging, the processor supports AI models including U-Net and Med-PaLM 2 for on-device 3D imaging and clinical reasoning.
This AI direction is fairly familiar by now, even for these niche products. AMD has been targeting ‘industrial mega-trends’ with its embedded lineup for at least three years, and this P100 expansion continues in that direction.
Open software, one hypervisor
ROCm support gives embedded developers access to an open-source AI stack, including HIP for hardware-agnostic GPU programming, without rewriting existing code. A Xen hypervisor-based virtualized reference stack allows Linux, Windows, Ubuntu, and RTOS environments to coexist in isolated domains on a single platform.