test

The Nx AI Manager integrates several runtimes that allow you to benefit from hardware acceleration when running deployed models.

Fully supported

The following AI accelerators are available out of the box in the current release:

AI Accelerator	Support Level	CPU Architecture	Operating System	API/driver version
CPU (OpenVINO)	Supported	x86_64	Windows, Linux	OpenVINO runtime (see OpenVINO page)
CPU	Supported	aarch64	Linux	-
Nvidia CUDA	Supported	x86_64	Windows	CUDA 11, CUDA 12
Nvidia Jetson Orin	Supported	aarch64	Linux (Jetson Orin)	JetPack (Orin supported versions)

Advanced configuration

These accelerators are supported through advanced configuration and additional validation. Contact Nx if you plan to use them in production.

AI Accelerator	Support Level	CPU Architecture	Operating System	API/driver version
Hailo (Hailo 8, Hailo 8L)	Advanced configuration	x86_64, aarch64	Linux	HailoRT (see Hailo Support page)
DeepX	Advanced configuration	x86_64, aarch64	Linux	DeepX SDK (vendor specific)

Requested and planned integrations

The following vendors have requested and committed to an integration with Nx AI Manager. These accelerators are not yet available in the current release.

AI Accelerator	Support Level	CPU Architecture	Operating System	API/driver version
EdgeCortix	Planned integration	TBD	TBD	TBD
Axelera	Planned integration	TBD	TBD	TBD
SiMa.ai	Planned integration	TBD	TBD	TBD
MemryX	Planned integration	TBD	TBD	TBD