An industrial mini PC watchdog is a chipset counter with a reset line: software writes to it on a schedule, and when the writes stop, the board restarts.
That sentence covers the whole idea, yet the feature is specified badly more often than almost any other line on an industrial datasheet. This is where the timer sits in an x86 machine, how it differs from the two features it gets confused with, and how to prove it works before a cabinet is closed.
1. Where the timer sits in an x86 mini PC
On an Intel platform the watchdog is normally part of the chipset rather than the processor. The timer lives in the TCO logic or in a Super I/O block, is enabled through firmware setup, and is programmed by the operating system through a driver. It keeps counting as long as the board has standby power, which is why it can issue a reset when the processor itself has stopped responding.
The GW-E specification lists this as watchdog programming with a hardware reset function, in 256 steps across 0 to 255 seconds. Those numbers describe what the register accepts, not what a site needs; the useful part is the word reset, because it tells you the signal reaches the board rather than only a process.
Software access differs by platform. On Linux, Intel chipsets are usually handled by the iTCO watchdog driver, and the state can be read and changed through the standard watchdog interface. On Windows the same hardware needs a vendor driver or a firmware-provided interface, so a quote that mentions watchdog support without naming the driver is only half an answer. On a fanless embedded computer the logic is identical: the chassis changes the thermal budget, not the reset path.
2. Watchdog, power-on restore and RTC schedule: three different jobs
Buyers routinely ask for one and mean another, and the confusion survives into the purchase order. The three features sit next to each other in a BIOS menu and solve separate problems.
| Feature | What it handles | What it does not |
|---|---|---|
| Watchdog timer | A running system that stops responding | A board that has lost power |
| Restore on AC loss | Power returning after an outage | A system that is on but frozen |
| RTC schedule | Switching on at a set hour | Recovery at any other time |
A control cabinet usually needs the first two together and the third only if the site runs shifts. The GW-E documentation lists power-on boot, scheduled boot and remote power on or off as separate options, which is the right way to present them: three switches, three behaviours, one line each on the order.
3. Hardware reset, NMI, or just restarting the application?
A watchdog can be wired to three different outcomes, and the difference decides what your operators see. A full hardware reset reboots the board: POST runs, the operating system starts again, and the machine is available in the time a normal boot takes. That is the behaviour most unattended sites want.
An NMI is a diagnostic path rather than a recovery path. It interrupts the processor so that a crash dump can be captured, which is useful during commissioning and almost useless in production, because the machine still needs a manual restart afterwards.
That combination, a service supervisor plus a board-level timer, is what separates an industrial pc solution from a desktop board in a metal box. The third option is a supervisor process that restarts a single service. It is fast and it keeps the rest of the system running, but it only works while the kernel and the scheduler are healthy. Pairing it with a hardware timer is the usual arrangement: the service restarts first, and the timer catches the case where the service cannot be restarted at all.
4. Choosing a timeout for a control cabinet
Start from the longest healthy cycle, not from the shortest number the register accepts. A machine that polls a PLC every second, writes a batch record every ten minutes and runs a database compaction at night has three different blocking times, and the watchdog has to tolerate the longest of them.
Then decide what a reset costs. On a line where a restart means a fifteen-minute warm-up and a recalibration, a 30-second timeout turns a brief glitch into a long stop. On a remote cabinet where nobody visits for a month, a short timeout is worth more than a graceful one. The number should be written down with the reason, because it is the setting most likely to be changed on site without a record.
Deployments in Berlin and London usually add a second constraint: the update window. Remote firmware and application updates can hold the system in a state that looks frozen for several minutes, so the timer has to be extended or disarmed for the duration and re-armed afterwards. If that path is not designed, the first update becomes a reboot loop across the estate.
5. What a watchdog cannot fix
It cannot cool a chassis, and it cannot replace a thermal design. A unit that shuts down because the cabinet has no airflow will keep shutting down, only now with a reboot between each event. The same applies to a supply that sags below the input range: a watchdog reset during a brown-out can leave the storage half-written, and no control cabinet industrial pc improves by resetting more often.
It cannot repair a worn SSD or a failing memory module, and it cannot catch an application that is wrong but still feeding the timer. That last case is the most common false sense of security: a loop that logs its heartbeat while returning stale data looks healthy to every watchdog ever built.
So treat the timer as one layer. Wide input voltage, a thermal path sized for the enclosure, a read-only system image and a health check that tests the real output are the layers around it. A specification that lists only the watchdog is a specification that has not thought about the other four.
6. A test plan you can run in an afternoon
Read the timer state first, then break it on purpose. On Linux, the standard watchdog interface reports whether the timer is active and how long is left; stop the feeding process and time the interval to reset. On Windows, the vendor driver or firmware utility provides the same control, and the system log afterwards should contain an unexpected shutdown entry rather than a clean restart.
Then repeat the test under load: a full processor, a busy storage device and a network cable pulled mid-transfer. Run at least thirty cycles, because a watchdog that works once has proved very little. Check that the file system comes back clean, that the application starts without manual input, and that the reset reason is recorded where a remote tool can read it.
Finish with the update path. Trigger a firmware update, interrupt it halfway, and confirm that the device either completes the update or returns to the previous image. That single test catches more field failures than any number of clean reboots.
7. What to ask before the order
◆ Is the timer in the chipset, in a Super I/O block, or on a separate IC?
◆ What is the timeout range and step size, and can it be changed from the operating system?
◆ Does expiry produce a full hardware reset, an NMI, or a service restart?
◆ Which driver is used on Linux and on Windows, and is it supplied?
◆ Can the timer be disarmed for updates and shutdown?
◆ Is the reset reason stored, and can it be read remotely?
◆ Does firmware setup expose power-on restore and scheduled boot as separate options?
◆ What test evidence comes with the sample?
The short version
In an x86 mini PC the watchdog is a chipset counter with a board-level reset, enabled in firmware and programmed by a driver. Specify the timeout by the longest healthy cycle, keep the recovery layers around it, and test it under load rather than on a bench. The feature is cheap; the difference between a working one and a datasheet line is the test report.
Where to look next
◆ The BVS GW-E industrial mini PC with its watchdog and BIOS options.
◆ Other chassis and port layouts in the industrial mini PC collection.
◆ Display and compute in one housing in the Windows panel PC range.