Case Study: AOI System Through Workload Consolidation Brings 100X Data Transmission Enhancement

  • June 21, 2022
  • Case Study
Case Study: AOI System Through Workload Consolidation Brings 100X Data Transmission Enhancement
Case Study: AOI System Through Workload Consolidation Brings 100X Data Transmission Enhancement

One leading DMS company used the iEPF-9010S Series through workload consolidation approach for phenomenal upgrades in its AOI system. One iEPF-9010S can process all AOI tasks for product quality inspection, replacing the traditional setting of using at least two IPC systems and improving overall images data transmission performance. The results show improved data transfer rate up to one hundred times for increased product inspection efficiency with fewer devices to manage, thus smaller system equipment footprint and lower system integration complexity.


A leading DMS company would like to build AOI systems with machine vision inspection technology to improve product quality and customer satisfaction. This DMS company has been using Windows OS-based industrial control PCs to run long-term machine automation. When importing product inspections, a Linux-based system with NVIDIA® GPU runs inspection models for image analysis.

Because the two operating systems are different, the company used two IPC systems to conduct product quality inspections for the entire AOI process. One industrial control PC runs on Windows OS to catch and process images from PoE camera whilst controlling the robotic arm. The other GPU computing system functions under Linux OS and works with NVIDIA® GPU to process AI inference and image analysis for product quality inspection. However, assessment through two IPC systems results in high latency due to images transmission through physical LAN. Hence, the AOI station becomes the bottleneck of the entire production process.


The data transmission speed between the two IPC systems significantly affects the efficiency of AOI inspection. To solve the data transmission latency problem between the two systems, ASRock Industrial provided the DMS company with one iEPF-9010S to perform all AOI tasks through the workload consolidation approach as the tailored solution. The iEPF-9010S system runs on Linux Ubuntu OS and sets up two virtual machines with KVM hypervisor. Virtual machine 1 (VM1) runs on Windows OS to grab the camera image and execute the motion control. In contrast, virtual machine 2 (VM2) runs on Linux OS to perform image analysis and AI inference. With the virtual machines set up, data transmission between both virtual machines through the virtual LAN results in a 10% increase in transmission speed compared to physical LAN between two systems. We further designed a software tool to utilize the shared memory for both virtual machines. It then efficiently expands the limit of transmission speed, significantly bringing up the transmission speed by 100 times compared to data transmitted by physical LAN in the traditional setting. 

SuperComputing power, rich IOs and expansions to power up AOI system

The iEPF-9010S Series features Intel 12th Generation Core Processor for powerful computing capability, rich I/Os, and flexible expansions to connect multiple industrial devices and empower AOI systems. Motion control is enabled through Motion Card to connect PLC and Motors to move inspection items in the conveyor belt of the production line. Through PoE connected to camera and light controllers, the AOI system takes photos of inspection items for defect detection running on the powerful Intel CPU with NVIDIA GPU card and shows real-time display of product inspection results through VGA port. After receiving the result, the controlling robotic arm transports the item to the next station. For other equipment maintenance capabilities, Digital Output ports link the tower light to indicate any abnormality in status/condition to signal maintenance alerts.

Workload consolidation and shared memory show 100X data transmission enhancement

The iEPF-9010S runs on Ubuntu OS and uses the KVM hypervisor to allocate the CPU cores and HW resources in the virtual machine. The first virtual machine (VM1), with the CPU Core 1-6 running on Windows OS functions as an industrial controller, executes motion control, and grabs images via PoE camera. With the CPU Core 7-8 running on Linux OS, the second virtual machine (VM2) performs the AI inference and image analysis with NVIDIA® GPU card for quality inspection. Data transmission between virtual machines through virtual LAN can increase the efficiency by 10% compared to using physical LAN between two systems.

However, the 10% data transmission enhancement is not enough to match with the other production module’s throughput, so the AOI station remains the bottleneck of the entire production process. To address this, we designed a customized software tool, leveraging the Ivshmem feature of KVM, to use the shared memory between two virtual machines. This tool provides seamless accessibility for both virtual machines. It significantly increases the image data transmission speed up to one hundred times faster than data transmitted by the physical LAN in the customer’s traditional setting. Through this solution, the AOI inspection station is no longer the bottleneck of the manufacturing process, providing higher customer satisfaction.


  • 100X data transmission enhancement: through workload consolidation and shared memory approaches, the new AOI system powered by iEPF-9010S Series reduces data transmission time and enhances speed to one hundred times faster than transmission by physical LAN for upgrades in overall product inspection efficiency.
  • Fewer devices to manage and lower system integration complexity: one iEPF-9010S for the AOI system replaces the traditional setting of using two IPCs, significantly reducing system, maintenance, and labor costs.
  • Smaller system equipment footprint: from traditional two IPCs to one iEPF-9010S, the single device solution occupies the smallest plant space possible, maximizing production line layout and ensuring space utilization at the most optimal level.

Did you enjoy this great article?

Check out our free e-newsletters to read more great articles..