Quality inspection virtual assistant with ADLINK COM-HPC, YOLOv8 & View AI WITH ADLINK COM-HPC-ALT SERVER
By Iain Menzies-Runciman
Introduction
Quality inspection is an essential part of any manufacturing process. For many industries, such as food packaging, AI vision systems can track objects and spot defects faster and with greater accuracy than traditional human inspection teams.
Recent AI innovations have also made it possible to take this raw data and generate instant natural-language insights. The result is increased production efficiency, as reports vital for system performance can be generated automatically.
Running AI vision and natural-language querying pipelines on the same hardware is a challenge, but not impossible. The ADLINK COM-HPC-ALT Size E server module provides an excellent platform, with lower power consumption compared to GPU-based AI.
This performance comes from the on-board Ampere Altra Max SoC, featuring 128 Arm CPU cores that can be partitioned into VMs. These VMs can run pipelines like Ultralytics YOLOv8 (real-time video analysis) and View AI (natural-language querying).
This guide covers the architecture, requirements, benefits, and a setup demonstration for real-time inferencing and querying on manufacturing video data.
AI-assisted manufacturing quality inspection system requirements
When integrating AI tools, manufacturing inspection systems require a sophisticated edge computer plus standard peripherals such as cameras, HMI devices, storage, and networking.
Figure 1. https://www.devheads.io/wp-content/uploads/2025/09/2.png.webp
We focus on the edge computer and SoC requirements for running two AI pipelines (YOLOv8 + View AI) in real time on separate VMs.
This subsystem must provide:
- Low power consumption for continuous AI workloads.
- Fast processing and high bandwidth to keep up with video analysis.
- Multi-core independent processing to run simultaneous AI workloads.
ADLINK COM-HPC-ALT server module
The ADLINK COM-HPC-ALT server module is based on the Ampere Altra family of multi-core Arm SoCs, ideal for edge and cloud-native applications requiring many CPU cores with predictable performance and low power.
Figure 2. https://www.devheads.io/wp-content/uploads/2025/09/3.png.webp
It uses the COM-HPC Server Type Size E Module form factor from PICMG. The open-standard design supports custom baseboards without vendor lock-in.
The integrated Ampere Altra SoCs feature 32–128 cores, supporting many VMs using Docker for containerized applications.
Ultralytics YOLOv8 provides high-speed object detection, classification, and tracking.
View AI enables natural-language querying of raw datasets from YOLOv8 through an easy chat interface.
The 128 cores of the Ampere Altra Max SoC
With 128 single-threaded Arm v8.2+ CPU cores, this is the largest option for the COM-HPC-ALT.
Figure 3. https://www.devheads.io/wp-content/uploads/2025/09/4.png.webp
The SoC can be divided into many VMs, supporting multiple AI pipelines. Each 64-bit core operates up to 3.0 GHz.
- Best price-to-performance for CPU-only AI — scalable, efficient, lower cost than GPU-based systems.
- Up to 768 GB DDR4 RAM — via 6× DIMM sockets at up to 3200 MT/s.
- 3 × 16 PCIe Gen4 lanes — minimizing bottlenecks for YOLOv8 and View AI.
Getting started walkthrough
Hardware setup
The Ampere Altra Dev Kit enables engineers to develop advanced AI workloads and connects to standard peripherals.
Figure 4. https://www.devheads.io/wp-content/uploads/2025/09/5.png.webp
Required components:
- USB keyboard, USB mouse, and VGA/HDMI display
- USB stick for OS flashing
- DB-9 cable + host machine
- Ethernet cable
- Approved RAM
- M.2 SSD
- PC power supply
Setup steps:
- Mount module and SSD
- Connect fan, heatsink, and RAM
- Connect Ethernet
- Connect HMI peripherals
- Connect power
Note: Preinstalled OS is Ubuntu 20.04, but the demo requires Ubuntu 24.04 ARM.
Installing Arm Ubuntu 24.04 OS
Steps:
- Create bootable USB with Ubuntu Server 24.04 for ARM64
- Connect DB-9 cable (VGA_COM0 → host)
- Install Minicom on host
- Identify device, launch Minicom
- Insert USB stick
- Power on the Ampere system to install OS
Setting up YOLOv8 and View AI demos
The demo performs real-time inferencing (YOLOv8) and natural-language querying (View AI).
YOLOv8 Demo Setup
- Enable firewall:
$ sudo apt update
2. Install Docker:
$ sudo apt-get update
3. Test Docker:
$ sudo docker run hello-world
- Add user to Docker:
$ sudo usermod -aG docker $USER
- Log out/in and verify:
$ groups
- Clone demo repo:
$ git clone https://github.com/ampere-solution/View-Ampere-AI-Demo.git
- Start demo:
./start-yolo-cpu.sh
- Open browser at http://:5002
Figure 5. https://www.devheads.io/wp-content/uploads/2025/09/6.png.webp
View AI Demo Setup
- Install QEMU/KVM:
$ sudo apt update
- Enable libvirt:
$ sudo systemctl enable --now libvirtd
- Add user to libvirt:
$ sudo usermod -aG libvirt $(whoami)
- Download Ubuntu 24.04 ARM ISO into:
/var/lib/libvirt/images/ - Create VM:
virt-install --name view-demo \
- Check VM:
$ virsh list -all
- Access VM:
$ virsh console view-demo
- Install View AI:
$ curl -s https://get.view.io | bash -s YOUR_GUID
Figure 6. https://www.devheads.io/wp-content/uploads/2025/09/7.png.webp
Here you can query quality-inspection feeds and create knowledge bases tailored to your audience.
Key takeaways
ADLINK’s COM-HPC-ALT module with the Ampere Altra Max SoC is a powerful platform for running YOLOv8 and View AI simultaneously.
Hardware advantages:
- Low power consumption (≈96 W for dual AI pipelines)
- Flexible VM-based architecture
- Standards-based COM-HPC design
Software advantages:
- YOLOv8: real-time performance
- YOLOv8: multi-object detection
- View AI: turns raw data into natural-language insights
The Ampere Altra Max SoC is an excellent platform for automated manufacturing inspection and multi-pipeline AI workloads.