A maker once disassembled a retired medical device from the 1990s and found a PC/104 industrial motherboard inside—slightly larger than a palm, with serial, parallel, and ISA buses built in. The board was transplanted into a handheld enclosure running Windows 98, and the whole machine kept working in a new role.
The same pattern shows up in factories: the enclosure, peripherals, sensors, PLC, and HMI still work, but the computer itself has reached end of life. Replacing the host alone is usually far cheaper than replacing the whole system. The right replacement here is an industrial arm computer—it takes over the computing role while the rest of the installation stays put.
Start with a triage. If two of three signals are present, the "swap the brain, keep the body" path is worth considering.
Peripherals still in service. PLCs, VFDs, sensors, and meters installed in the last decade, with stable models and ample spares. The wiring, terminal blocks, and mounting hardware are all proven in the field.
Only the computer is at end of life. BIOS no longer updated, IDE connectors with no replacement drives, an operating system that no longer receives security patches—yet the PLC program still uploads and downloads, and sensor readings still report correctly. The fault sits entirely in the host, not in the attached devices.
Budget is tight. A full replacement means cabinet redesign, rewiring, HMI reconfiguration, and a week of production downtime. In most cases the total runs 5 to 10 times the cost of swapping the host alone.
Path A: keep the enclosure, power supply, and peripheral cabling, and replace only the motherboard. Remove the old ATX or PC/104 board and fit a small-form-factor ARM industrial PC (industrial arm computer). The key step is interface matching: RS-485 connects to the PLC, RS-232 to legacy instruments, CAN to the VFDs, and Ethernet to the HMI. Physical connector positions need not match exactly, but interface types and port counts must cover the originals. If the legacy system used a USB keyboard, mouse, and VGA display, the new board should retain HDMI or VGA accordingly.
Path B: keep the original system as the running core, and hang an ARM industrial PC alongside as an edge node. The original system keeps running the critical process; the ARM host attaches over Ethernet or serial and takes on data acquisition, protocol conversion, cloud reporting, and local caching. The two machines communicate over Modbus TCP or OPC UA, and any fault on the new node can be rolled back without touching the original process. This dual-machine arrangement is safer than a hard cutover.
For sites mixing legacy and new interfaces, Path A is more direct. For systems running critical processes with high downtime risk, Path B is the safer choice.
Interface density. RS-485, RS-232, CAN, DI/DO, Ethernet, and USB all need to be present in sufficient numbers. The worst outcome is a new board with fewer interfaces than the old one, which forces a pile of external converters back into the cabinet.
Long Linux lifecycle. Ubuntu LTS (22.04, 24.04) ships five years of security updates, and applications can migrate across generations. ARM platforms such as the RK3562, RK3568, and RK3588 families carry 10+ year supply commitments from the silicon vendors—a meaningful factor for sites that plan to run for a decade.
Wide voltage, wide temperature, and local SCADA. Unstable site power and cabinet temperatures above 50°C are common in industrial environments. DC 9-36V input, -25 to +75°C operating range, and a dual hardware/software watchdog are baseline requirements. A built-in Local SCADA/HMI engine that handles data acquisition, screen configuration, and reporting cuts the engineering effort sharply, since no separate HMI panel needs to be added.
Certification continuity. CE, ROHS, and 3C are hard requirements for industrial devices. After a host swap, the whole unit typically needs recertification; choosing a model that already holds the relevant certifications avoids one extra certification cycle.
For medium interface density, a Linux software stack, and fast onboarding, USR-EG528 is a sensible starting point: an RK3562 quad-core Cortex-A53 running at 2.0GHz, Ubuntu 24.04 with Node-RED and the WukongEdge engine, 4× RS485 (2 switchable to RS485/232), 2× 100M Ethernet, 2× USB, HDMI and audio, DC 9-36V, -25 to +75°C, fanless aluminum housing, certified 3C, CE, ROHS, WEEE. The platform, interfaces, and software stack all carry across generations, which is exactly what a "swap the brain, keep the body" upgrade needs.
For denser interfaces with multiple meters and digital I/O, plus some edge AI, USR-EG828-GL offers an ARM Cortex-A55 quad-core at 2.0GHz with a 1TOPS NPU, RS232/RS485/CAN, AI/DI/DO, WiFi/4G/eSIM, optional Linux Ubuntu 20.04 or Android, certified CE, ROHS, WEEE. Multi-protocol acquisition, rule logic, and local linkage run in one box, removing the need for an external serial server.
For vision AI with high compute demand, USR-EG928A uses an RK3588J octa-core 64-bit with a 6TOPS NPU and ARM Mali-G610 GPU, Ubuntu 22.04, 8K encode/decode, gigabit Ethernet, Wi-Fi 6, 5G/4G expansion, -25 to +75°C, certified 3C, CE, FCC, ROHS, WEEE. It belongs to the new Edge AI Box class and fits best as a node on the vision-AI upgrade path.
Upgrading a legacy industrial PC is essentially a triage: whether the peripherals and PLC program are still in service decides if the host alone can be swapped, while interface types, software stack, power and temperature, and certification continuity decide whether the new ARM host will hold up. The value of an industrial arm computer in this generation is not that ARM beats x86, but that ARM industrial PCs have been industrialized enough on interface density, platform longevity, local configuration, and wide voltage and temperature to fit the "swap the brain, keep the body" upgrade path. Keeping the surrounding system
