Occasional and erratic freezing?
Occasional and erratic freezing?
I possess a compact headless PC housed in a cabinet within a dry, cold environment. This machine operates continuously as a NAS unit and has increasingly taken on the task of handling CPU-heavy computations that my slow laptop couldn't manage. It constantly works hard, keeping all CPU threads at full capacity.
Since its setup many years ago, it has experienced sporadic, unpredictable freezes. It functions well for some periods before suddenly becoming completely unresponsive. There are no SSH replies, the router no longer recognizes it, and connecting a monitor and keyboard only displays code page boot info and login prompts without any cursor movement. The NumLock indicator on the keyboard fails to react to key presses. The system remains stuck, with attempts like Ctrl+Alt+Delete or SysRq+FKREISUB yielding no effect.
Over time, I've attempted various fixes: BIOS updates, switching between operating systems (from Windows 10 to Debian, Arch, FreeBSD, Obarun), running Memtest86+, replacing RAM, and reconfiguring components. Despite these efforts, the freezes persist.
In the BIOS, I adjusted numerous settings—voltage levels, clock speeds, power-saving features, threading, and more—yet nothing changed.
I've also run monitoring tools that capture snapshots every minute: CPU temperature, disk status, running processes, available RAM, and events leading up to crashes. Sadly, these snapshots only reflect a stable system operating smoothly.
I suspect the issue may lie among three possible areas:
- The power supply unit
- The central processing unit
- The motherboard
Given my budget constraints, I prefer not to replace these parts. Is there any method to identify which component is causing the problem?
PSU: model, wattage, age, condition (original to build, new, used, refurbished)?
Have you inspected the electrical socket and the circuit that powers the PC/NAS?
Are there any UPSs, surge protectors, power strips, extension cords, etc.?
Has the CMOS battery been swapped for a brand new one?
"Cold dry room" - static buildup and discharge? That would probably be random and could cause all sorts of problems
PSU: Corsair VS350, arrived new in 2021
Electrical socket dedicated, direct connection to fuse, 2.5 mm², 24h probe recorded average 230V (min 229V, max 230V)
No surge protector or UPS used; only the cord provided with the PSU was connected between socket and unit
CMOS battery updated in 2023
The environment is cold and dry, preventing condensation. The hygrometer indicates a humidity level of 45% today.
45% sits right around the middle between low humidity (40%) and high humidity (50%). If possible, adjusting it to 50% or 55% might help. Refer to the PC's documentation for suitable ranges or settings. It could be worth a test.
PSU is running continuously now and is 4 to 5 years old. This would be my initial attempt to replace it.
Make sure to back up the server first and use only the cables that come with the new PSU.
Could you share your budget range so I can suggest suitable PSUs?
Purchase a PSU replacement from a store offering favorable return terms.
Seasonic has consistently been reliable for me.
Check out the Seasonic GX-650 with a 7-year warranty:
https://www.amazon.com/Seasonic-CORE-V2-...M9VWD?th=1
I bought and set up a Seasonic G12 GC650, and also swapped out the CMOS battery once more for a new one. The old battery still worked fine. However, the freezing issue remains unchanged.
While reviewing old logs, I observed something peculiar: whenever the machine freezes, the RTC loses time noticeably. Openntpd gradually reduces the offset, but once a freeze occurs, the RTC becomes off by hundreds or thousands of seconds in a negative direction. At one point, I thought freezes occurred only when the RTC reached zero offset, but that turned out to be incorrect. Is this a sign anyone should be aware of?
Any advice on what to do next would be appreciated.
Since the issue has persisted after multiple changes—replacing the CMOS battery twice, updating BIOS, reseating components, and switching OSes—I believe this points to a hardware problem, likely with the motherboard, and replacing it would be the best solution.
The new PSU was a worthwhile investment because you definitely needed one.