Preempt RT Linux on TX2- Are these results normal for RT ?

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi,

I follow the instruction here for Preempt RT Linux https://github.com/kozyilmaz/nvidia-jetson-rt .

I used hackbench test for CPU load and test latency with Cyclictest.

For 5M samples with %50 CPU Load

CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
Min Latencies: 00005 00007 00007 00005 00005 00005
Avg Latencies: 00009 00014 00014 00009 00009 00009
Max Latencies: 00095 00112 00125 00091 00061 00078

For 10M samples with %100 CPU Load
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
Min Latencies: 00006 00008 00008 00006 00005 00005
Avg Latencies: 00011 00018 00018 00011 00011 00011
Max Latencies: 00338 00555 00191 00357 00176 00346

Latencies in microseconds.

Are these results normal ? I think Max Latencies are high for a real time operating system. Also I wonder why is that the denver cores worse than the ARM cortex core at almost all tests that I made ? I did my all test with nvpmodel -m 0 and jetsonclock.

Thanks.