Compute next to the process, and keep the data on site
The SimtoLink edge computing box is an industrial design with a multi-core CPU. It runs Docker and K3s (lightweight Kubernetes) containers, so data collection, AI inference, and video analysis can sit on the box. That is compute at the edge, which takes load off the cloud and cuts how much the site depends on the network.
Use it where the data has to stay on site: the edge of a production line, an equipment room, or a store. The SimtoLink edge runtime is already installed, so the box is ready out of the carton. It is already in electronics, automotive, retail, and energy, and it speaks several industrial protocols and AI frameworks.
Compute options and how to choose
The box is offered in several compute builds, matched to the application load. The table is a typical set, not a full catalog:
| Build | Typical specification | Typical use |
|---|---|---|
| Base | 4-core CPU, 8GB RAM, 128GB SSD | Data collection, protocol conversion, light applications |
| Standard | 8-core CPU, 16–32GB RAM | Several containers, video-stream processing |
| GPU | NVIDIA Jetson / Intel integrated graphics | AI inference, object detection, image classification |
| Industrial wide-temperature | Extended operating temperature, dust and shock | Outdoors, high temperature, vibration |
Interfaces are several Ethernet ports, USB, and HDMI. An external disk can extend storage, and a 19-inch rack kit is available. Match the compute build to the application load; SimtoLink supplies a performance test and a selection tool. On the GPU build, watch the cooling, and derate the unit in high temperature.
Where it runs, and what changed
From AOI on an SMT line, to plate recognition at a filling station, to store analytics in retail, the box covers compute that has to stay close to the work. These are typical jobs and what they changed on site:
AOI defect detection on an SMT line
An electronics plant put the box on the SMT line to run an AOI defect model and watch yield in real time. The data stays on the line, so the cloud is not in the path, which is what a low-latency, compliance-bound line needs.
Plate recognition and nozzle data at a filling station
A filling station runs plate recognition and fuel-nozzle collection on the box. The site depends less on the central equipment room, and each station can manage itself and send a summary upstream.
Footfall and shelf analysis in a store
A supermarket chain runs footfall counts and shelf recognition on a box in each store. The video is processed locally and only the summary goes up, which keeps privacy tighter and the bandwidth bill down.
What to settle before you install
- Application: Take the base build for data collection alone. Take the GPU build when you need AI inference. Take the large-memory build when several containers share the box.
- Site: Outdoors or in high temperature, take the industrial wide-temperature build. Before you order, be clear on the application, the compute you need, and the site.
- Isolation: Set CPU and memory limits on each container so one application cannot take the box down.
- Edge apps: SimtoLink keeps an edge application store, so data collection, AI inference, and video analysis can go on quickly.
- Custom load: A named software set can be preinstalled. SimtoLink also supplies a performance test and a selection tool.
Rollout and looking after the box
A practical order for putting an edge computing box in:
Remote deployment and OTA updates keep the maintenance load down. The edge application store and SimtoLink support cover the common apps. On the GPU build, watch the cooling, and derate the unit in high temperature.
Support
SimtoLink helps with selection and with getting applications onto the box, including AI inference and video analysis, plus compute sizing and performance tuning. A product manual and a project quote are available.