Geniatech AIM-B2 Edge AI Module Delivers 40 TOPS with PCIe Gen4

Geniatech AIM-B2 NXP AI Accelerator 

Geniatech has introduced the AIM-B2, a compact NXP AI Accelerator module powered by the NXP Ara240 processor with Kinara Ara-2 edge AI architecture. The module targets edge AI, generative AI, robotics, industrial automation, and smart city applications. It combines the NXP Ara240 AI processor with a PCIe Gen4 interface in a compact 40 × 55 mm form factor, designed for systems that require high AI throughput with low power consumption.

The AIM-B2 delivers up to 40 TOPS INT8 AI performance while consuming around 12W of power. Geniatech designed the module for workloads including object detection, semantic segmentation, pose estimation, facial recognition, and transformer-based AI models. The platform also supports multi-chip scalability, allowing developers to expand AI processing performance for larger edge deployments.

Geniatech AIM-B2 Top

Key Features of the AIM-B2 NXP AI Accelerator:

• Processor: Kinara Ara-2 edge AI processor with up to 40 TOPS INT8 AI performance
• Memory: Up to 16GB LPDDR4X memory with 4GB and 8GB options
• Host Interface: 4-lane PCIe Gen4 interface through B2B connector
• Generative AI Support: Supports LLaMA 2 and Stable Diffusion models
• AI Performance: ResNet50 inference latency as low as 2 ms
• Stable Diffusion: Stable Diffusion 1.4 generation in around 10 seconds for 20 iterations
• Host Platform Compatibility: Supports NXP, NVIDIA, Qualcomm, and Xilinx platforms
• AI Functions: Supports object detection, semantic segmentation, pose estimation, and facial recognition
• Scalability: Multi-chip support for scalable AI workloads
• Connectivity: PCIe, USB, UART, and I2C interfaces exposed through dual B2B connectors
• Power Consumption: Typical power consumption rated at 12W
• Operating Temperature: 0°C to 70°C industrial operating range
• Dimensions: 40 × 55 mm module size with 1.2 mm PCB thickness

The module exposes PCIe Gen4 lanes, USB signals, UART, I2C, reference clocks, and control interfaces through dual B2B connectors. This layout allows system integrators to embed the accelerator into robotics controllers, AI gateways, industrial PCs, smart cameras, and autonomous drone systems. The design focuses on stable high-speed connectivity and long-term industrial deployment.

Geniatech AIM-B2 Bottom

The AIM-B2 supports Linux and Windows operating systems and works with major AI frameworks including TensorFlow, TorchScript, PyTorch, ONNX, Caffe, and MXNet. Developers can deploy computer vision, transformer, and generative AI models without major framework conversion issues. Geniatech also provides SDKs, optimized drivers, and integration support to simplify deployment across industrial and embedded platforms.

Geniatech has listed the AIM-B2 AI accelerator module on its official website, but the company has not published pricing yet. Additional details, documentation, and purchase inquiries remain available through the official product page at Geniatech AIM‑B2 product page.

Images used courtesy of Geniatech.