Is the built-in watchdog in the chip reliable?

[tilblackout]

We hope that the system can be reset promptly when the program crashes, so we use a watchdog to detect these situations.
However, I vaguely heard some people say that the built-in watchdog in the chip may not be reliable.
But I think those big chip manufacturers are not that foolish,could the program freezing also cause the internal watchdog of the chip to crash?

In some products I have developed, hardware engineers deliberately connected a TPL5010 watchdog chip(TPL5010 chip reference manual:https://www.ti.com/lit/ds/symlink/tpl5010.pdf) to the reset pin of the MCU (such as STM32, I.MX RT1170, Kinetis MK64).

I don't believe that the internal watchdogs of these chips are unreliable. What is your opinion? Otherwise, what are the application scenarios for watchdog chips like TPL5010?

[RoGeorge]

What products are you saying about?  You have to be more specific, or else we can only speculate. 

My guess is that that design choice is not because an external watchdog would be more reliable than the internal one.  An external watchdog might be more reliable, for example if the external watchdog is rad-hard while the microcontroller is not.  But I don't think that was the case here.

Some other times, there are design requests based on laws, or based on regulations that are mandatory.  To give an example, some industrial installations in power plants and energy distribution must be redundant.  And for a better resilience against unknown future pitfalls, the redundancy must be implemented with different brands.

Or, maybe the external watchdog was added mostly to lower the power consumption, and that would be the most reasonable guess.  The external timer chip is supposed to be even lower power than the standby consumption of the MCU with its internal watchdog still running.  So, the external watchdog-timer here is rather used like a wake-up timer, and not like a watchdog that resets the MCU from a malfunction.

[PCB.Wiz]

We hope that the system can be reset promptly when the program crashes, so we use a watchdog to detect these situations.
However, I vaguely heard some people say that the built-in watchdog in the chip may not be reliable.
But I think those big chip manufacturers are not that foolish,could the program freezing also cause the internal watchdog of the chip to crash?

In some products I have developed, hardware engineers deliberately connected a TPL5010 watchdog chip(TPL5010 chip reference manual:https://www.ti.com/lit/ds/symlink/tpl5010.pdf) to the reset pin of the MCU (such as STM32, I.MX RT1170, Kinetis MK64).

I don't believe that the internal watchdogs of these chips are unreliable. What is your opinion? Otherwise, what are the application scenarios for watchdog chips like TPL5010?

There are many terms in play.
Watchdog oscillators are very reliable, but proper watchdog operation is certainly non-trivial. It should involve testing by forcing erratic operation.

Many novices start the watchdog, and start an interrupt to service the watchdog.     Job Done ?

Some industry sectors have preference for a power removal watchdog. That gives a full power cycle restart, as there are many failure cases where hitting the MCU-RST alone is not enough.

Complex operating system watchdogs often have dual or multiple timeouts. They allow a longer boot-up time, before they apply a shorter service time.

The TI part is interesting, it is very low power and has surprising accuracy for such a low power, but it also has a power-up phase, where it measures the set-point.