I want to run some compute and I/O intensive P2P software. However, no matter what I do the system can’t run stable for more than 30 minutes. I have also tried copying data from another machine using rsync
. That also leads to a crash eventually (<30 minutes).
My hardware config:
- Radxa Rock 5B
- Radxa power adapter (I tried a variety of others but bought the official one eventually just to be sure)
- WD Black SN850X 4TB NVME
- OS running on microSD (I tried running the OS from NVME as well, doesn’t make a difference)
I tried various OS images:
- rock-5b_debian_bullseye_cli_b39
- rock-5b-archlinux-installer
- Armbian_23.2.1_Rock-5b_jammy_legacy_5.10.160 (<-- most stable so far)
Observations:
- Power meter at power socket seems to suggest power consumption usually doesn’t exceed 12W
- CPU and NVME temperature doesn’t usually exceed 70˚C (I have also tried throttling the system and maintaining <40˚C as well but that doesn’t help either)
- Not running anything (idle system) seems ‘stable’
Tried (without effect?):
- Throttling mem clock:
echo 1560000000 | sudo tee /sys/class/devfreq/dmc/max_freq
- Throttling CPU clock:
echo 408000 | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_setspeed
- Setting affinity of my processes to ensure only 4 out 8 cores are being used heavily to reduce overall load
No matter what I do the system eventually crashes. Symptoms vary. Sometimes it’s a core dump. Often the OS becomes corrupted/unusable: for example, my P2P processes are still running alongside other things such as htop
but entering new commands in the terminal result in ‘Segmentation Fault’ or other weird errors (e.g., shared libraries are corrupted etc…). Shortly after the whole system crashes.
I am really frustrated right now and could use some help.