PSU problem?

[MacCheetah3]

My SFF rig has started exhibiting a problem. I’m not quite sure where to start. Admittedly, I haven’t done much troubleshooting yet. A quick visual inspection did not reveal anything obvious.

When I press the power button, the power LED flashes for a fraction of a second. Any subsequent press results in no activity, although the GPU ‘standby’ power LED does remain illuminated, which is normal. The solution thus far has been to unplug from AC for 30 seconds or so, plug back in, then the system appears to function fine. No other ill signs, even under heavier load. I thought it might have been a glitch but the behavior is consistent.

The PC is connected to a (less than year old) UPS. There have been several power disturbances/outages this summer. Even with the UPS, could the PC’s PSU have been damaged? Maybe coincidentally failing PSU capacitors?

I think the PSU is this model:

SF Series SF750 Fully Modular 80 PLUS Platinum SFX Power Supply

Outfit your next SFF PC with a CORSAIR SF Series 80 PLUS Platinum SFX Power Supply for compact, ATX 3.1 compliant, fully modular Platinum power.

www.corsair.com

The PC was originally an Origin PC build. However, I’ve upgraded/swapped the platform and GPU since.

Beyond the inconvenience, I’m sincerely concerned this could be a deeper problem. And… Well… for obvious (current economic) reasons I don’t want the possibility of (other) components to be taken down. Thus, foremost, despite it seemingly working fine on second boot, should I not run it for lengthy periods without narrowing down the cause?

I’ll try some further troubleshooting though… even reseating RAM feels scary as it’s probably more valuable per ounce than gold right now. 😆 🙁

 

[diamond.g]

Case power switch bridges pin 16+17 on the 24 pin motherboard connector. If the power supply is now requiring a hard reset (either via the plug or its physical switch) that sounds like something is tripping its internal protections.

Does it power off in the middle of being used?

I’m not sure if I would advise opening the power supply to take a look. You may be able to get a replacement from Corsair since these have a 7 year warranty.

 

[MacCheetah3]

Here’s a visual:

Normal boot:

Another observation: The PC powers (back) on fine if shut down within the past 30 minutes. I will leave it off for a little over an hour — if I don’t forget to check soon after that. Basically, I will try to find the threshold.

EDIT: Two hours later (i.e., now)… Powered on normal. 🤔

Does it power off in the middle of being used?

Not thus far.

You may be able to get a replacement from Corsair since these have a 7 year warranty.

I’m doubting even with Origin PC being a subsidiary of Corsair. I think the SI’s warranty takes precedence.

 

[sgtaylor5]

I would find an inexpensive ATX power supply tester on Amazon first. If there is a bad power lane, then it will flash and the tester will beep.

 

[MacCheetah3]

I would find an inexpensive ATX power supply tester on Amazon first. If there is a bad power lane, then it will flash and the tester will beep.

Not a bad idea. It's not as if the PSU needs to be under load to trigger. Actually, the issue appears to be just the initial spike and probably a fault on the 12V rail.

I had the trip once happen coming out of sleep (not hibernate).* And I've found that if I discharge things first, disconnect the AC, and press the power button before trying an actual power on, there seems to be no problem.

Definitely a PSU issue I have not seen before. For me, they've either worked or full on fail.

* That phrase could be very misconstrued out of context. 😂

 

[MacCheetah3]

I may have confirmed my initial assumption. 🙁

This behavior indicates a latched protection circuit or failing capacitors within the PSU. The unit is not simply "discharging" in a normal way; it is failing to reset its internal safety logic until the residual energy in its capacitors fully dissipates.

The Mechanism: Latched Protection​

Modern PSU units use latch-style protection circuits (for Over Current, Over Voltage, or Short Circuit). Unlike "hiccup mode" (which rapidly retries power), a latch keeps the unit disabled until the main internal capacitors discharge below a specific threshold (often ~20V).

• Why the Delay?: The "30-minute" wait you described is the time required for the bulk capacitors to naturally bleed off their stored charge through internal resistors. If these resistors are weak or the capacitors are degraded, this discharge takes much longer than the typical few seconds.
• The "Click": The sound you hear is the protection relay engaging. When you try to power on immediately, the control chip detects the latch is still active (because voltage remains on the capacitors) and instantly trips the relay again.
• Why It Works Later: Once unplugged for 30 minutes, the voltage drops low enough to reset the latch logic, allowing the PSU to attempt a fresh startup sequence.

Likely Causes​

1. Failing Electrolytic Capacitors: As capacitors age, they can lose capacitance or develop high leakage, causing them to hold charge longer or behave erratically during the startup phase. This is a common failure mode in older units.
2. Intermittent Short Circuit: There may be a component inside the PSU (or a connected component like the motherboard) that develops a short only when warm or under specific electrical conditions. The PSU detects this, trips, and locks out until fully reset.
3. Faulty Protection Circuitry: The monitoring chip itself may be defective, triggering a false positive that requires a full power cycle to clear.

Recommendation​

Replace the PSU immediately. While the unit currently works after a long wait, this symptom is a classic precursor to total failure. The underlying issue (bad capacitors or unstable protection logic) will likely worsen, leading to:

• Inability to start even after waiting.
• Sudden shutdowns under load.
• Potential damage to other components (motherboard, GPU) if the protection fails to trip correctly in the future.

Do not rely on a PSU exhibiting "latch-up" behavior; it is effectively defective.

I probed, queried 'search AI' further -- had to call out some hallucinating logic along the way, of course:

You are absolutely correct to challenge that logic. A "cold" boot actually creates a HIGHER electrical stress event than a warm boot, which explains why your failing PSU trips on Linux but seemingly worked (or failed differently) on Windows.

Here is the corrected technical explanation for why the Linux cold boot triggers the protection trip while the Windows warm state did not:

1. Inrush Current: The "Cold" Shock​

• Cold Boot (Linux): When a PSU is fully discharged (unplugged for 30 mins), its internal bulk capacitors are empty. Upon pressing the power button, the PSU must draw a massive surge of current (Inrush Current) from the wall outlet to charge these capacitors instantly while simultaneously trying to power the motherboard.
  • The Failure: If your PSU’s Inrush Current Limiter (usually an NTC thermistor) or input rectifiers are degraded, this massive initial spike triggers the Over Current Protection (OCP) or Short Circuit Protection (SCP) immediately. The PSU detects a "short" because the demand to charge empty caps looks like a dead short to a failing circuit.
• Warm/Soft Boot (Windows): Windows (even with Fast Startup disabled) often leaves the PSU in a "standby" state where capacitors retain some charge, or it performs a reboot sequence where capacitors never fully discharge.
  • The Benefit: Because the capacitors are already partially charged, the Inrush Current required to start the system is significantly lower. This lower stress level stays just below the threshold of your failing protection circuit, allowing the system to boot.

2. The "Discharge" Paradox​

You mentioned the PSU works after being unplugged for 30 minutes. This is not because it "needs" to discharge, but because:

1. Latch Reset: The 30-minute wait allows the voltage to drop low enough to reset the latched protection IC.
2. The Trade-off: Once reset, the next boot is a Cold Boot.
   • If the PSU is severely degraded, even the reset isn't enough; the inrush current trips it again (your current Linux experience).
   • If the degradation is moderate, the reset allows one successful boot, but subsequent reboots (warm) work fine because they avoid the high inrush stress.

3. Why Linux Exposed It​

• Windows Behavior: You likely performed Restarts or Shutdowns that didn't fully drain the capacitors (due to motherboard leakage or standby rails), keeping the system in a "Warm" state where inrush current was low.
• Linux Behavior: Pop!_OS likely performs a stricter ACPI S5 shutdown, cutting all power rails cleanly. This ensures the capacitors fully discharge. When you boot back up, it is a true Cold Boot with maximum inrush current, which immediately trips your weakened protection circuit.

Conclusion​

Your intuition was right: The cold boot is the harder event.

• Windows was inadvertently "babying" the failing PSU by keeping it warm (low inrush).
• Linux is forcing a true cold boot (high inrush), which the dying PSU can no longer handle.

Final Verdict: The PSU has a failing input stage (NTC thermistor, bridge rectifier, or bulk capacitors) that cannot handle the inrush current of a cold start. Replace the PSU immediately; it is now failing even the basic startup sequence.

If anybody can vet this, it would be helpful.

Nonetheless... The fix appears to be replacing the PSU. The PC being an SFF rig makes such a task far more tedious, which is mainly why I've procrastinated on that route.

Sad 🐼*

P.S. If I need to do a full tear down anyway, I'm tempted to look/try a different SFF case. EDIT: The NCASE M2 looks quite nice aesthetically and functionally, the "Grater" version specifically.