Skip to main content

NXP Ara240 M.2 Module

Discrete M.2 AI accelerator module delivering 40 eTOPS for plug-in generative AI at the edge

40 eTOPS | 16GB LPDDR4 onboard | M.2 2280 M-Key | PCIe Gen4 x4 | ~6.6W

View on NXP.com

Overview

The NXP Ara240 M.2 module is a discrete neural processing unit (DNPU) packaged in the standard M.2 2280 M-Key form factor, delivering 40 eTOPS of AI acceleration via PCIe Gen4 x4. With 16GB of onboard LPDDR4 memory, it can run models up to 30 billion parameters (INT4) independently of host memory. At approximately 6.6W typical power, it enables fanless industrial deployments.

Specifications

SpecificationValueNotes
NPUNXP Ara240 (multiple NPU + VPU cores)
AI Performance40 eTOPSINT4, INT8, mixed precision
Onboard Memory16GB LPDDR4
Host InterfacePCIe Gen4 x4 (~8 GB/s)
Form FactorM.2 2280 M-Key (NGFF)
Power~6.6W typical
Max Model SizeUp to 30B parameters (INT4)
FrameworksTensorFlow, PyTorch, ONNX
SoftwareNXP Ara SDK
StatusPreproduction
CompatibilityFRDM i.MX 8M Plus, FRDM i.MX 95, other M.2 M-Key hosts

Use Cases

  • Adding generative AI (LLM/VLM) inference to existing embedded systems via standard M.2 slot
  • Accelerating computer vision and neural network workloads on NXP i.MX or other host platforms
  • Low-power edge AI inference in industrial, medical, and retail devices without requiring a GPU