WOLF-3570 XMC GPU and video capture module
EMBEDDED GPUs and EDGE AI
Embedded systems · Edge AI

Image Processing and Edge AI with WOLF

Wolf Advanced Technology develops embedded computing modules that combine NVIDIA graphics processing units (GPUs), video capture technologies and high-speed networking for image processing and AI applications. Its XMC, 3UVPX, 6UVPX, and VNX+ options address different integration requirements. Video processing, network-to-GPU data transfer and computing in limited space are three key areas of the portfolio.

RequirementProcessing data at its source
TechnologyGPU, video and data paths
Local supportFrom requirements to solutions with TARGET
WOLF product family

Product portfolio for a range of requirements

Three modules in the WOLF portfolio address video processing, high-speed networking and compact edge AI requirements.

WOLF-3570 XMC module
XMC

WOLF-3570

Combines an NVIDIA RTX™ 2000 Ada GPU and FGX2 video capture on one XMC module. With 8 GB of GDDR6 memory and configuration-dependent SDI inputs and outputs, it is designed to add GPU and video I/O capabilities to an existing carrier.

WOLF-153L 3U VPX module
3U VPX

WOLF-153L

Combines an NVIDIA RTX™ 5000 Ada GPU, 16 GB of GDDR6 memory and a ConnectX-7 SmartNIC on a 3U VPX module. With configurable networking up to 100 GbE and GPUDirect RDMA support, it is designed for network-to-GPU data transfer.

WOLF-N4XP VNX+ module
VNX+

WOLF-N4XP

Provides the NVIDIA® Jetson Orin™ NX 16 GB platform in a compact VNX+ form factor. Camera, PCIe, Ethernet and storage connections make it an option for edge AI applications with limited installation space.

VNX+ · N4XP

Edge AI in a compact footprint

N4XP is a Jetson-based computing module for teams looking to process image and sensor data close to its source. It combines a CPU, GPU, memory and NVMe storage in a compact VNX+ module. For machine vision and AI inference, it can support software that runs on the device without sending all data to a remote server. A 1000BASE-T Ethernet connection links the module to the local network, while onboard storage allows data to be retained locally.

MIPI CSI-2 camera connectivity and WOLF software support based on Jetson Linux and JetPack provide a foundation for integrating cameras with applications. Module selection should consider the camera interface, software version, memory requirements and space available for cooling.

N4XP VNX+ computing module based on NVIDIA® Jetson Orin™ NX
N4XPNVIDIA® Jetson Orin™ NX · VNX+

Configuration and thermal design depend on the target application's operating conditions.

From requirements to architecture

Explore the right architecture for your application

Capturing camera data, transferring it to a GPU over a network or running local analysis: each workflow calls for a different combination of connectivity and computing.

Select your application requirement

Video from an SDI camera can be captured by the FGX2 on WOLF-3570 and processed by the GPU. The application sends the image analysis or video processing output through the selected interface. Planning starts with camera count, resolution, frame rate and input/output formats. Integration also requires consideration of the XMC carrier's PCIe connection, video signal routing and driver support. Delivery of processed video through SDI or a suitable display interface depends on the selected configuration.

InputSDI camera
WOLF-3570FGX2 + GPU
OutputVideo
Conceptual data flow

For data arriving over Ethernet, WOLF-153L combines a GPU and ConnectX-7 on one module. With a suitable system architecture, GPUDirect RDMA is intended to reduce data copies through the host processor when transferring data to GPU memory. Network bandwidth, data volume, VPX backplane connections and software support should be planned together. Results can be shared over the network as the application requires. The 100 GbE figure is the maximum link rate; application performance is measured using real data.

InputEthernet
WOLF-153LConnectX-7 + GPU
OutputNetwork
Conceptual data flow

N4XP can run image analysis and AI inference close to the system collecting the data. This approach can be considered for machine vision, local video analysis and sensor data preprocessing. Selection starts with the AI model's memory needs, camera compatibility and processing time targets. Power, cooling and storage requirements also form part of the compact system design. Where local recording is needed, data generation rate and retention time should guide storage planning.

InputCamera / sensor
N4XPJetson Orin™ NX
OutputLocal analysis
Conceptual data flow

TARGET support

TARGET is WOLF’s partner in Türkiye. TARGET supports product selection, technical pre-assessment and project coordination for projects in Türkiye.

Contact TARGET