Dear NVIDIA team，
We have a custom carrier board+Orin 32GB module that cannot be started at high temperatures. The specific description is as follows:
- Set the temperature box to 70 ℃ and execute the command in the system: .
/gpu_burn 20000; - Using JTOP to view information, it was found that when “loc1” exceeded 106 ℃, the device was physically powered off, waited for 1 minute, and then powered on again. The system could not start, and there was no printing information on the Orin debugging port.
At first, we thought it was caused by the high temperature of “loc1” (which is the internal temperature of I350), so our hardware engineer removed the I350 chip and continued the experiment, but the problem still existed. From here, it can be tentatively determined that it is not related to I350.
Later, we measured the signals related to the power on timing and found that the CARRIER_PWR_ON and SYS-REET-N signals were low. Therefore, our Orin onboard microcontroller (responsible for executing the Orin power on timing) added detection to these two signals after completing the power on timing. If they were low, the power on timing was re executed. After verification, it was found that the microcontroller executed n (n>1) times before the module started.
The module can start 100% normally at room temperature, but cannot start 100% after high temperature. Because there is no printing information for the Orin module’s debugging serial port, I currently don’t know where to troubleshoot.
Can you give me some advice?
When we power off the device, the JTOP interface information is as follows:
