Edge devices in industrial settings need to run VPN tunnels, handle protocol conversion, and sometimes support secondary development. Choosing between a purpose-built industrial router and an industrial computer running OpenWrt is a decision many engineers face. Three decision criteria help clarify it.
An industrial router is a finished solution delivered by the vendor: hardware, operating system, network protocol stack, and remote management platform are already integrated, and the unit works once powered on.
An industrial PC running OpenWrt is a self-assembled approach: buy a general-purpose industrial computer, flash an open-source routing system onto it, and configure networking manually. The flexibility is real, but so is the technical debt.
OpenWrt itself is mature, with an active community. In industrial settings, though, "it runs" and "it runs reliably for years" are different questions. The latter tends to be resolved not at the network layer but at the levels of environmental tolerance and day-to-day operations.
The most common failure mode on industrial sites is not weak performance. It is going offline while nobody is watching.
Consider a pump house. Equipment sits in an outdoor distribution cabinet where internal temperature reaches 65℃ in summer and drops below freezing in winter, powered by 24V DC with noticeable fluctuation. What such an environment demands is wide temperature and voltage tolerance, a metal enclosure, a hardware watchdog, and DIN rail mounting, plus automatic redial after link loss, SIM failover, and traffic threshold alerts.
A properly specified industrial router ships all of that as standard. The USR-G806w matches this profile with a Qualcomm platform, a -20℃ to +70℃ operating range, hardware and software watchdogs, surge and ESD protection, both DIN rail and wall mounting, and built-in network anomaly alerts.
With an industrial PC running OpenWrt, these same items become engineering tasks:
On a staffed site with regular inspection rounds, these can be tuned gradually. For unattended sites or service points scattered across multiple cities, out-of-the-box reliability outweighs any hardware price gap.
The test is straightforward: what else must the device do besides networking?
When the answer covers VPN, firewall, port forwarding, and link backup only, an industrial router is sufficient.
When the answer includes a local database, Docker containers, video AI inference, PLC control logic, or protocol conversion across CAN bus and multiple RS485 ports, an industrial computer enters the picture.
The USR-EG628 takes on this latter class of work as an ARM-based industrial computer: RK3562J industrial-grade chip, quad-core Cortex-A53 at up to 2.0GHz, Ubuntu preinstalled, Node-RED and Docker supported, with 2 RS485, 1 RS232, 1 CAN, and 2 Ethernet ports onboard, and capable of PLC programming and video AI workloads.
In short: networking-centric work points to a router, compute-centric work points to an industrial computer. When both are needed, splitting responsibilities works well — the router handles connectivity, the industrial computer handles computation, and the two connect over Ethernet.
The reliability of cellular connectivity is the dividing line between "integrated by default" and "assembled in-house."
In an industrial router, the 4G/5G module arrives fully adapted: dialing scripts, signal monitoring, SIM switching, module reset, and band compatibility have all been validated across many field deployments.
On an industrial PC running OpenWrt, all of that is hand-built: USB mode switching, AT command debugging, driver compilation, firmware upgrades. Change the module model and the work repeats itself. A discontinued module forces another round.
The arithmetic runs like this: hardware cost may favor the industrial PC, but the integration cost of bringing up a cellular module and configuring failover, combined with the ongoing maintenance cost of firmware updates, security patches, troubleshooting, and spare parts management, overtakes the hardware savings once deployment passes a few dozen units. In the other direction, for a handful of units in a lab doing validation work, the flexibility of the industrial PC route is genuinely worth more.
Remote management deserves mention too. With several hundred units distributed across regions, servicing them through on-site serial cables is unrealistic. The management platform bundled with industrial routers supports remote configuration, firmware upgrades, and status monitoring, capabilities that cost considerably more to build independently.
This question tends to get tangled up with the one above. Sorting it out by data volume:
The USR-G816 fits 5G deployments: Qualcomm quad-core processor with an X62 modem, SA+NSA dual mode, -35℃ to +75℃ operating range, serial and multiple Ethernet ports, and full VPN protocol support including OpenVPN, IPsec, and GRE. That -35℃ low-temperature rating is a hard requirement for outdoor projects in northern climates.
Working through it in order:
Most industrial sites land on an industrial router. The industrial computer earns its place where edge compute is genuinely required. When the answer remains unclear, requesting an evaluation unit and testing it on site against real workloads, actual temperatures, and live signal conditions tends to settle the question. Measured data says more than a datasheet.
1. What actually separates an industrial router from a consumer router?
Not CPU clock speed or wireless throughput, but environmental tolerance and reliability engineering. Concretely: wide temperature ratings of the -35℃ to +75℃ class, wide-voltage input, a metal enclosure, DIN rail mounting, surge and ESD protection, a hardware watchdog, automatic recovery after link loss, SIM failover, and a remote management platform. Put a consumer router in an industrial cabinet and the usual outcome is not weak performance but heat-related lockups or failure to recover after a power event.
2. For a remote site, is 5G the right move or does 4G still hold up?
Two factors matter: data volume and site lifespan. Low-volume work such as PLC polling and telemetry runs fine on 4G, with lower module cost and cheaper data plans. Video backhaul, bulk synchronization, or sites planned for 8-10 years favor 5G, since operators gradually refarm 4G spectrum and long-running projects need to account for network longevity. Standalone mode matters as well. Latency-sensitive applications such as remote real-time control gain little from NSA, which anchors to the 4G core.
3. Does an industrial router on site need Wi-Fi?
Often not. Many industrial routers sit inside cabinets with equipment connected by Ethernet or serial, where Wi-Fi only widens the attack surface and is better disabled. Wi-Fi proves useful where mobile assets such as AGVs or inspection devices need access, or where a technician arriving on site needs a temporary debugging channel. Where Wi-Fi is required, checking whether antennas are detachable, whether higher-gain external antennas can be fitted, and whether SSID broadcast can be turned off is worth doing.
4. What is the correct way to reach a PLC behind a router remotely?
Have the router initiate the connection outward, either through a VPN tunnel or over a private APN from the carrier. Mapping PLC ports directly to a public IP address exposes control equipment to constant scanning from the internet, which is a live risk rather than a theoretical one. Once the tunnel is established, the control room reaches the PLC through private addressing inside the tunnel, with no open ports visible from the public side. This is precisely why IPsec, OpenVPN, and GRE come standard on industrial routers.
5. Can a router recover on its own after losing connectivity?
A well-built industrial router can. The escalation typically runs from a connectivity check that probes whether a target address is actually reachable rather than reading signal bars, to a cellular module reset, then failover to the backup SIM, and only then a full reboot. A hardware watchdog serves as the final backstop against software hangs. Worth confirming whether this behavior is merely "supported" or is enabled by default, and whether the interval at each escalation step is configurable.
6. Is one SIM enough, and when does dual SIM become necessary?
One SIM covers sites with stable coverage and reliable operator service. Dual SIM earns consideration for unattended sites in remote areas, mobile installations in vehicles or vessels, and situations where downtime carries a high cost, such as payment terminals or security monitoring. The underlying value lies in not staking reliability on one operator's network quality. Note the difference between dual SIM standby failover and dual SIM active load balancing. Hardware implementation and tariff structures differ between them.
7. How should cellular data usage be estimated and kept under control?
Start from actual per-unit load: PLC-class telemetry typically runs hundreds of bytes per point at second- or minute-level polling intervals; a single 1080P video stream generally consumes 2-4 Mbps by bitrate; remote desktop bursts depend on the specific operation. Multiply by unit count and a redundancy factor. On the control side, router-side traffic accounting and threshold alerting are essential. Industrial routers commonly support monthly thresholds with email or SMS alerts as the limit approaches, and can throttle or cut the link outright. Provisioning at roughly 1.5 times the estimate initially and adjusting to measured usage after a month works well.
8. What certification requirements apply to export projects?
Mandatory certifications vary by region. The EU requires CE marking, and since 2024 the cybersecurity requirements under the RED directive (EN 18031) have been mandatory for wireless and cellular equipment. The US requires FCC, China requires CCC, and markets such as Australia (RCM), Brazil (ANATEL), and Thailand (NBTC) each maintain their own requirements. Requesting certificates specific to the target market, with certificate numbers and a
