What defines edge AI performance today? A compact workstation integrates CPU, GPU, and NPU to deliver powerful local processing for demanding applications.
ASRock Industrial has introduced the AI BOX-A395, a compact AI workstation designed to deliver high-performance local processing for data-intensive workloads. Built on AMD Ryzen AI Max+ 395 processors, the system integrates CPU, GPU, and NPU capabilities to enable advanced AI applications directly at the edge.
As AI deployment expands across industries, the need to process data closer to its source is becoming more critical. Cloud-based models often introduce latency, higher operational costs, and data security concerns. The AI BOX-A395 addresses these limitations by enabling on-device AI inference, allowing applications such as large language models, generative AI, and vision analytics to run locally with faster response times and improved data control.
At its core, the system features an APU architecture with up to 16 CPU cores and 32 threads, combined with integrated graphics and an AI engine delivering up to 50 TOPS of acceleration. This unified design supports a wide range of workloads, including AI development, engineering simulations, and high-resolution content creation, all within a compact footprint.
To handle memory-intensive tasks, the platform supports up to 128GB of LPDDR5x unified memory, enabling large models to run without external accelerators. Dual M.2 slots with PCIe Gen4 support and RAID configurations provide scalable storage, while Wi-Fi 7 and Bluetooth 5.4 ensure high-speed connectivity.
The system also includes USB4, 10GbE networking, and multi-display support with HDMI 2.1 and DisplayPort 2.1, enabling high-bandwidth data transfer and up to 8K output. A multi-heat pipe thermal design ensures stable performance under sustained workloads, while enterprise features such as TPM 2.0 and redundant BIOS support enhance reliability.
By combining compute density with edge-focused design, the system reflects the growing shift toward localized, efficient AI processing.
Click here for the official announcement.