High temperature shutdown on TX2i

We use a Tegra TX2i card on a custom carrier card. The set is integrated into a box (aluminum) with a heat sink.
When we do temperature tests, the Tegra TX2i card seems to fall into an idel state that arrives at 52°C.
At the time of losing external communication (Ethernet, UART,…), the internal temperatures given by the Tegra are as follows:
• BCPU-therm-virtual-0: 74.5°C
• MCPU-therm-virtual-0: 74.5°C
• GPU-therm-virtual-0: 72.5°C
• Tboard-tegra-virtual-0: 71°C
• Tdiode-tegra-virtual-0: 72.5°C
• thermal-fan-is-virtual-0: 73.7°C
The temperature on the tegra heat sink is 72.3°C.
When we lose the connection with the Tegra, the overall consumption of the equipment decreases from 24Watt to 12Watt.
We can see that the power-carrier signal switches between 0 and 1 quickly (~10Hz).
When we trying to control the Tegra’s reset signal, we can’t restart it. The only way is to completely turn off the power and restart the equipment.
We did a dozen tests, the cut always occurs around 52°C ambient.
Do you have any idea of the origin of the cut or a track to investigate?
Thank you

Hi, it is strange as temperature range of TX2i is -40C to 85C which was validated before ship out. Have you done cross check test on dev kit? How many modules have you tested? Did you record the thermal zones value just at the moment of failure happen or there is a continuous record till failure happen?

Hi Trumany,

We have done continuous record during the test. The values given are the last before the Tegra shutdown.
We have make the same test with 2 tegra with the same results but not with the dev kit.
We have also tested restart (by reset pin pulse) after the decreasing of the temperature, but without effects.
Can you say me which drive the CARRIER_PWR_ON signal on the tegra and why the signal switch between high and low level ?

Thanks for your help.

Can you do the test on devkit to confirm if module has problem? CARRIER_PWR_ON is from module to indicate power status which allow carrier board to power on, switching to high/low level is due to system status.

Also please share the log file here to help check.

Any news about this issue?

Sorry for the delay.
We have resolve the problem.
It is not a Tegra problem, but a problem on the the CPLD which drive the reset signal of the Tegra on my custom board.