When the system is under full load, if we first stress the serial port, network port, GPU and memory, and then use the stress command to load all CPU cores, the 1st CPU core will show a load of less than 1% (while all other CPU cores can be loaded above 99%).
Verification steps:
Check with the top and jtop commands. Under this scenario, the load on the 1st CPU core is less than 1%;
Use the stress command to load only all CPU cores independently; all cores can be loaded above 99%.
This issue is reproducible most of the time. Occasionally, the CPU1 utilization in jtop will suddenly spike from very low to a high level, matching that of the other CPU cores, and it appears to return to normal completely. What could be the root cause? I would appreciate any help. Thank you very much.
Which version of Jetpack release is in use here?
want to clarify. So your test is based on power mode 10W?
How is the situation when using other power mode?
Could you test with 40W mode instead of maxN and share us the result?
Your previous screenshot also only shows 10W so I don’t know if that one has anything valuable to check.
Retested: one NX and one Nano, iperf3 send/receive on 4 network ports,then stress all the CPU
So why the CPU1 utilization shown in jtop is extremely low?
Why does CPU1 return to normal when iperf3 is stopped and only stress is running?
Nano:
iperf3+stress:
only stress:
NX:
iperf3+stress:
only stress:
So what are the exact steps to see this behavior? Sounds like iperf plays a role here.
What is the interface that iperf is testing? The on-board ethernet? or something else?
