40 TOPS M.2 AI Accelerator Module

AIM-M-K

40 TOPS M.2 AI Accelerator Module

Geniatech AIM-M-K is a M.2 AI accelerator powered by the NXP Ara240 NPU. Leveraging the standard M.2 M-key interface, it enables seamless integration into edge servers, embedded systems, and industrial PCs. Optimized for transformers and generative AI models, the AIM-M-K supports LLaMA 2.0, YOLOv8, and major AI frameworks such as ONNX, TensorFlow, and PyTorch. It is ideal for edge AI deployment.

40 TOPS Performance: Powered by NXP Ara240 NPU
M.2 Interface: Easy integration into industrial and embedded platforms
AI framework compatibility with TensorFlow, PyTorch, ONNX
Optimized for transformer and generative AI models
Supports LLaMA 2.0, YOLOv8, and advanced AI models
Supports concurrent multi-stream and multi-model inferencing
Fast inference: Stable Diffusion 1.4 ~10s (20 iterations), ResNet50 2ms latency

Chipset & NPU	NXP Ara-240
Arithmetic power	40 TOPS
RAM	16GB (8GB Optional)
PCB Thickness / Size	0.9mm / 22 × 80mm
AI Frameworks	TensorFlow, PyTorch, ONNX
Performance	Stable Diffusion 1.4: 7 sec/image Llama2-7B: 12 output tokens/sec MobileNetV1 SSD: 974 IPS (1.03 ms latency)
Security	Secure Boot, Root-of-trust processor, encrypted interface
Host Processor	x86, Arm
Interface	M.2 (M-KEY) / 4 Lane PCIe Gen4
Operating Voltage	DC 3.3V
Power	TDP: 12W / Typical: 3W / Full: 8W
Cooling	Heatsink
Temperature	Operating: 0°C – 50°C / Storage: -40°C – 85°C
Humidity	Operating: 5% – 90% / Storage: 5% – 95%
OS Support	Linux & Windows
Certification	CE / FCC Class B (pending)
Market Area	Global