AIM-B2

40 TOPS AI Inference Accelerator with Kinara Ara-2

Geniatech AIM‑B2 is a compact, high‑performance AI acceleration module powered by the Kinara Ara‑2 processor. It delivers up to 40 TOPS of INT8 performance at just 12W, enabling low‑latency, energy‑efficient AI inference for generative AI and demanding edge computing applications.

  • Powered by Kinara Ara-2 with 40 TOPS edge AI performance
  • Supports GenAI models like LLaMA 2 and Stable Diffusion
  • Compatible with TensorFlow, PyTorch, ONNX, Caffe, and more
  • High-speed 4-lane PCIe Gen4 interface
  • Ultra-low latency: ResNet50 in 2ms, Stable Diffusion in 10s

Balancing Power & Cost

Powerful Edge AI

Powered by Kinara Ara‑2, the AIM‑B2 delivers up to 40 TOPS INT8 at only 12W. It balances high throughput with efficiency for edge and generative AI workloads.

Easy Integration

This compact (40 × 55 mm), production‑ready module embeds easily into existing designs. It significantly cuts development time and engineering complexity.

Broad AI Framework

It supports leading frameworks like TensorFlow, PyTorch, ONNX, Caffe, and MXNet. You can seamlessly deploy models from YOLOv8 and LLaMA 2 to Stable Diffusion.

Ultra‑Low Latency

Capable of processing ResNet50 in 2 ms and Stable Diffusion in 10 seconds, it delivers real‑time inference critical for vision, language, and generative AI applications at the edge.

Flexible Connectivity

Equipped with a high‑speed PCIe Gen4 ×4 interface, it ensures rapid data transfer and seamless integration with a wide range of host platforms such as NXP, NVIDIA, Qualcomm, and Xilinx.

Cost‑Effective Deployment

The modular design streamlines the Bill of Materials (BOM) and engineering effort, effectively lowering overall system costs and accelerating time‑to‑market for scalable deployments.

Reliable Performance

AIM‑B2 ensures consistent performance and long‑term stability in industrial settings, with up to 16 GB LPDDR4X memory and a 0 °C – 70 °C operating range

Complete Solution

Provided as a full package with optimized drivers, SDKs, and toolchains. Geniatech further accelerates your project with expert customization support and integration services.

Chip Kinara Ara-2
AI performance 40 TOPS, Support llama2.0 and Yolov8
Host Processor NXP,Nvidia,Qualcomm,Xilinx
AI Model Frameworks Supported TensorFlow,Torchscript, PyTorch&ONNX, Caffe, Mxnet
Memory Up to 16GB LPDDR4(X)
Host Interface B2B(4-lane PCIE Gen4)
OS Support Linux, Windows
Power Consumption(Typical) 12W
Operating Temperature 0℃~70℃
Dimensions 40×55mm

AIM-B2 Specification

Order Inquiry

Looking for volume pricing or ready to accelerate your project? Connect us for a detailed consultation.