www.ptreview.co.uk
27
'26
Written on Modified on
ARBOR Technology Unveils COMX C710 AI Module For Next Generation Robotics
Powered by AMD Ryzen AI Embedded X100 processors, this COM-HPC module delivers high-performance heterogeneous computing for industrial automation, medical, and autonomous systems.
www.arbor-technology.com
ARBOR Technology has introduced the COMX-C710 computer-on-module, utilizing AMD Ryzen AI Embedded X100 processors to address processing demands in physical artificial intelligence. The hardware platform targets applications across autonomous mobile robots, medical imaging systems, and industrial automation control units.
Architecture and Hardware Specification
The COMX-C710 adheres to the COM-HPC Client Size C form factor standards, measuring 120 mm by 160 mm. The integrated processor platform delivers up to 126 TOPS of total compute performance for artificial intelligence workloads. This performance budget includes a dedicated neural processing unit capable of 50 TOPS based on the XDNA 2 architecture.
High-density LPDDR5x memory is integrated directly into the module architecture to maximize data transfer rates between processing units and minimize latency during high-throughput inference tasks. To accommodate harsh operating environments, the hardware includes support for wide-input voltage ranges and extended operating temperature thresholds.
Task Allocation in Heterogeneous Computing Systems
The platform distributes computing workloads across specialized hardware components to optimize energy efficiency and processing speed:
- The central processing unit manages deterministic operational logic, system control, and high-level decision algorithms.
- The graphics processing unit accelerates multi-camera visual pipelines, real-time spatial mapping, and simultaneous localization and mapping (SLAM) calculations.
- The neural processing unit processes object detection, classification, and scene parsing models, including YOLO neural network architectures.
Sumit Shah, head of product management and marketing in the Adaptive and Embedded Computing Group at AMD, stated that combining the CPU, GPU, and NPU on a single silicon die provides low-latency processing for edge automation platforms. Vincent Liao, Vice President of the Embedded Computing Product Division at ARBOR Technology, noted that consolidating heterogeneous compute units on standardized module form factors reduces system design complexity in medical and robotic equipment.
Target Applications and Industrial Deployment Options
The hardware design combines real-time motion control, image acquisition, and neural inference onto a single compute module. Target deployment environments include autonomous mobile robots, collaborative robotic arms, industrial automation gateways, high-resolution medical imaging devices, endoscopes, and patient monitoring systems. By consolidating multiple compute subsystems into a standardized module interface, deployment complexity is reduced across multi-sensor edge installations.
Additional Context
This section details technical specifications and competitive benchmarking not included in the original product announcement.
The COM-HPC Client Size C form factor specification defines high-speed interconnect pinouts, enabling support for high-lane-count PCIe channels and high-bandwidth memory buses compared to older legacy standards such as COM Express Type 6. In physical AI applications, integrated neural processing units compete with discrete PCIe accelerator cards. Embedded platforms with integrated NPU architectures achieve lower power consumption metrics per TOPS by avoiding external bus transfer overheads and utilizing low-power memory interfaces like LPDDR5x. Standard benchmark criteria for embedded modules in this class focus on inference throughput (frames per second per watt) on standardized models such as YOLOv8, memory bandwidth utilization, and thermal dissipation stability within fanless enclosures.
Edited by Evgeny Churilov, Induportals Media - Adapted by AI.
www.arbor-technology.com
Target Applications and Industrial Deployment Options
The hardware design combines real-time motion control, image acquisition, and neural inference onto a single compute module. Target deployment environments include autonomous mobile robots, collaborative robotic arms, industrial automation gateways, high-resolution medical imaging devices, endoscopes, and patient monitoring systems. By consolidating multiple compute subsystems into a standardized module interface, deployment complexity is reduced across multi-sensor edge installations.
Additional Context
This section details technical specifications and competitive benchmarking not included in the original product announcement.
The COM-HPC Client Size C form factor specification defines high-speed interconnect pinouts, enabling support for high-lane-count PCIe channels and high-bandwidth memory buses compared to older legacy standards such as COM Express Type 6. In physical AI applications, integrated neural processing units compete with discrete PCIe accelerator cards. Embedded platforms with integrated NPU architectures achieve lower power consumption metrics per TOPS by avoiding external bus transfer overheads and utilizing low-power memory interfaces like LPDDR5x. Standard benchmark criteria for embedded modules in this class focus on inference throughput (frames per second per watt) on standardized models such as YOLOv8, memory bandwidth utilization, and thermal dissipation stability within fanless enclosures.
Edited by Evgeny Churilov, Induportals Media - Adapted by AI.
www.arbor-technology.com
