Power Supply Diagnostics: Detecting PSU Problems Before They Cause Damage

Goal: Diagnose power supply issues on Windows and Linux using software analysis, voltage monitoring, and hardware testing Level: Intermediate Language: English

Why PSU Problems Are Hard to Detect

PSU failures are notoriously difficult to detect with software alone -- they often masquerade as disk errors, random crashes, or memory corruption. This section covers what software can reveal and when hardware testing is necessary.

Unlike disks (SMART), CPUs (thermal sensors), or RAM (ECC), power supplies have no standardized self-reporting mechanism. There is no "PSU health" attribute. Instead, we rely on:

Motherboard voltage sensors — most boards monitor the main ATX rails
Crash pattern analysis — the Windows Kernel-Power Event 41 contains clues
Symptom correlation — certain failure patterns point to power issues
Stress testing — pushing the PSU to its limits to trigger failures
Hardware measurement — multimeter or dedicated PSU tester

Symptoms That Point to PSU Problems

Strong Indicators

Symptom	Why It Points to PSU
System shuts off instantly (no BSOD, no warning)	CPU lost power — no time for a crash dump
Crashes only under heavy load (gaming, rendering)	PSU cannot sustain peak wattage
Multiple different BSOD codes (not the same one repeating)	Unstable voltages cause unpredictable failures
USB devices disconnect randomly	+5V rail sagging
SATA CRC errors on multiple drives simultaneously	+5V or +3.3V rail instability affects SATA signaling
System won't wake from sleep reliably	+5VSB (standby) rail is weak
Burning smell from the back of the case	Capacitor failure — shut down immediately

Weak Indicators (Could Be PSU, But Often Aren't)

Symptom	More Likely Cause
Single BSOD code repeating	Driver or hardware fault (RAM, disk)
CRC errors on only one drive	SATA cable for that specific drive
System runs fine under load but crashes at idle	Sleep/wake driver bug
Fan noise changes	Normal thermal management

Windows: Software-Based PSU Diagnostics

Step 1: Analyze Kernel-Power Event 41

Event 41 is logged every time Windows starts after an unclean shutdown. The BugcheckCode field inside the event XML reveals whether it was a software crash or a power loss.

$startDate = (Get-Date).AddDays(-30)
$kp41 = Get-WinEvent -FilterHashtable @{
    LogName      = 'System'
    ProviderName = 'Microsoft-Windows-Kernel-Power'
    Id           = 41
    StartTime    = $startDate
} -ErrorAction SilentlyContinue

if ($kp41) {
    Write-Host "Found $($kp41.Count) unexpected shutdown(s)`n"
    foreach ($ev in $kp41) {
        $xml = [xml]$ev.ToXml()
        $eventData = $xml.Event.EventData.Data
        $bugcheckCode = ($eventData | Where-Object { $_.Name -eq "BugcheckCode" }).'#text'
        $sleepInProgress = ($eventData | Where-Object { $_.Name -eq "SleepInProgress" }).'#text'

        Write-Host "  Time:            $($ev.TimeCreated)"
        Write-Host "  BugcheckCode:    $bugcheckCode"
        Write-Host "  SleepInProgress: $sleepInProgress"

        if ($bugcheckCode -eq "0") {
            Write-Host "  => POWER LOSS (no crash dump generated)" -ForegroundColor Red
        } else {
            Write-Host "  => Software crash (BSOD code: $bugcheckCode)"
        }
        Write-Host ""
    }
}

Interpreting Event 41 Fields

Field	Value	Meaning
BugcheckCode	0	No BSOD occurred — the system lost power instantly. This is the strongest software indicator of a PSU problem (or external power loss).
BugcheckCode	nonzero	A BSOD occurred — this is a software/driver/hardware crash, not necessarily PSU-related.
SleepInProgress	0	System was fully running when the event occurred
SleepInProgress	nonzero	System was in or transitioning to/from a sleep state

BugcheckCode = 0 + SleepInProgress = 6

This combination means the system lost power while sleeping. Common causes:

PSU's +5VSB (standby) rail is failing
Power strip or UPS cut power during sleep
BIOS sleep/wake settings are misconfigured
The PSU cannot maintain standby power for connected USB devices

Real-World Example

In the scenario that prompted this guide, 3 out of 5 unexpected shutdowns had BugcheckCode=0 with SleepInProgress=6 — the system was losing power during sleep. Combined with SATA CRC errors on multiple drives, this pointed to either PSU instability or external power interruptions.

Step 2: Read Motherboard Voltage Sensors

Windows has no built-in way to read voltage sensors. You need one of these tools:

Option A: HWiNFO64 (Recommended)

Download HWiNFO64 -- free, portable version available. For detailed remote monitoring and Grafana integration, see HWiNFO Remote Monitoring.

Run HWiNFO64 in Sensors-only mode
Find the motherboard chip section (typically Nuvoton, ITE, or Fintek)
Look for voltage readings: +3.3V, +5V, +12V, VBat, +5VSB
Leave it running for several hours and check Min/Max columns
Compare against the ATX specification (see reference table below)

HWiNFO64 can also log sensor data to CSV for long-term monitoring.

Option B: LibreHardwareMonitor (Scriptable)

LibreHardwareMonitor exposes a WMI namespace that PowerShell can query:

Download and run LibreHardwareMonitor
Enable Options > Remote Web Server or just leave it running
Use this PowerShell snippet to read voltages:

Get-WmiObject -Namespace root\LibreHardwareMonitor -Class Sensor |
    Where-Object { $_.SensorType -eq "Voltage" } |
    ForEach-Object {
        $val = [math]::Round($_.Value, 3)
        $min = [math]::Round($_.Min, 3)
        $max = [math]::Round($_.Max, 3)
        Write-Host "$($_.Name): ${val}V  (min: ${min}V, max: ${max}V)"
    }

Step 3: Stress Test the PSU

OCCT (free for personal use) has a dedicated PSU test mode:

Run OCCT and select the Power Supply test
This stresses CPU and GPU simultaneously to maximize power draw
Monitor voltages during the test (OCCT shows them in real-time)
If the system crashes during this test, the PSU is the most likely cause
Run for at least 30 minutes — some PSU failures only appear after thermal buildup

Before Running a PSU Stress Test

Save all work and close all applications
A failing PSU under stress can damage components
If you hear clicking, buzzing, or smell burning — stop immediately
Have a fire extinguisher rated for electrical fires accessible (not water-based)

Linux: Software-Based PSU Diagnostics

Step 1: Read Voltage Sensors with lm-sensors

# Install
sudo apt install lm-sensors    # Debian/Ubuntu
sudo dnf install lm_sensors    # RHEL/Fedora

# Detect sensors
sudo sensors-detect
# Answer YES to all probes (or review each one)

# Read current values
sensors

Example output:

nct6776-isa-0290
Adapter: ISA adapter
Vcore:           +0.90 V  (min =  +0.00 V, max =  +1.74 V)
in1:             +1.02 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
AVCC:            +3.34 V  (min =  +2.98 V, max =  +3.63 V)
+3.3V:           +3.38 V  (min =  +2.98 V, max =  +3.63 V)
in4:             +1.01 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
in5:             +1.26 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
in6:             +0.06 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
3VSB:            +3.42 V  (min =  +2.98 V, max =  +3.63 V)
Vbat:            +3.26 V  (min =  +2.70 V, max =  +3.63 V)
in9:            +1.728 V  (min =  +0.00 V, max =  +0.00 V)  ALARM

Ignore the ALARM Labels on Unused Inputs

Many motherboard sensor chips have inputs that are not connected to anything. These show 0V with ALARM status. Only pay attention to the labeled rails (+3.3V, +5V, +12V, VBat, 3VSB).

Step 2: Continuous Voltage Monitoring

# Log voltage readings every 60 seconds
while true; do
    echo "$(date '+%Y-%m-%d %H:%M:%S')" >> /tmp/voltage_log.txt
    sensors | grep -E '\+3\.3V|\+5V|\+12V|3VSB|Vbat' >> /tmp/voltage_log.txt
    echo "---" >> /tmp/voltage_log.txt
    sleep 60
done

Or use collectd or telegraf for proper time-series monitoring.

Step 3: Check for Power Loss Events

# Unexpected shutdowns (no clean shutdown logged before reboot)
journalctl --list-boots | head -20

# Look for boots where the previous session has no shutdown entry
# A missing "Stopped target" or "Reached target Shutdown" before a new boot = power loss

# Check for ACPI power events
journalctl -k | grep -iE 'ACPI.*power|power.supply|battery|critical'

# Check for hardware watchdog resets
journalctl -k | grep -i watchdog

# Voltage-related kernel messages
dmesg | grep -iE 'voltage|power|acpi|thermal|throttl'

# CPU throttling due to power issues
grep -r "cpu_frequency" /sys/devices/system/cpu/ 2>/dev/null

Step 5: Stress Test with stress-ng

# Install
sudo apt install stress-ng

# CPU + memory stress (maximizes power draw)
stress-ng --cpu $(nproc) --vm 2 --vm-bytes 1G --timeout 600 --metrics-brief

# Monitor voltages during the test in another terminal
watch -n 2 sensors

For GPU stress, use gpu-burn (NVIDIA) or furmark under Wine.

ATX Power Supply Voltage Specifications

Tolerance Table

Rail	Nominal	Min (-5%)	Max (+5%)	Danger Zone
+3.3V	3.300V	3.135V	3.465V	Below 3.1V or above 3.5V
+5V	5.000V	4.750V	5.250V	Below 4.7V or above 5.3V
+12V	12.000V	11.400V	12.600V	Below 11.0V or above 13.0V
-12V	-12.000V	-12.600V	-11.400V	Rarely monitored; legacy rail
+5VSB	5.000V	4.750V	5.250V	Standby rail — powers wake-on-LAN, USB charging

What Each Rail Powers

Rail	Components
+12V	CPU (via VRM), GPU (via PCIe power), fans, HDDs (motor), optical drives
+5V	SATA signaling, USB ports, some logic chips
+3.3V	RAM, chipset, SATA controller logic, some SSD internals
+5VSB	Wake-on-LAN, USB power in sleep, BIOS RTC, power button circuit

Voltage Deviations and Their Symptoms

Deviation	Typical Symptoms
+12V sag under load	System shuts off during gaming/rendering; GPU crashes
+12V ripple (fluctuation)	Random BSODs, memory errors, CPU throttling
+5V sag	USB device disconnections, SATA CRC errors, disk I/O failures
+3.3V sag	Memory errors, chipset instability, boot failures
+5VSB drops during sleep	System won't wake from sleep; power loss during standby

SATA CRC Errors and the +5V Rail

SATA data signaling operates on the +5V rail. If the PSU's +5V output is unstable or sagging under load, it can cause CRC errors on multiple SATA drives simultaneously — mimicking cable failures. If CRC errors appear on drives connected to different cables and different SATA ports, suspect the PSU's +5V rail before replacing cables.

Hardware Testing

When software diagnostics are inconclusive, measure voltages directly.

PSU Tester (Simple, ~10-15 EUR)

A dedicated PSU tester plugs into the 24-pin ATX connector and displays voltages on an LCD. Quick and easy, but only tests at zero load (no components drawing power).

Multimeter (More Accurate)

Measure voltages at the Molex or SATA power connectors while the system is running:

Pin	Wire Color	Expected Voltage
1 (Molex)	Yellow	+12V
2 (Molex)	Black	Ground
3 (Molex)	Black	Ground
4 (Molex)	Red	+5V

SATA Power Connector (15-pin):

Pins	Wire Color	Voltage
1-3	Orange	+3.3V
4-6	Black	Ground
7-9	Red	+5V
10-12	Black	Ground
13-15	Yellow	+12V

Safety

Never open the PSU case — capacitors inside can hold lethal charge even when unplugged
Measure at the connectors, never inside the unit
Use one hand when probing, keep the other hand away from the case
Set the multimeter to DC voltage, 20V range

When to Replace the PSU

Condition	Action
Any voltage outside the +/-5% ATX specification under load	Replace
BugcheckCode=0 crashes that stop after PSU replacement	Confirmed PSU failure
Visible capacitor bulging or leaking on the PSU's PCB (if visible through the fan grill)	Replace immediately
PSU is 7+ years old with any power-related symptoms	Preventive replacement recommended
System crashes only during high power draw (stress test, gaming)	PSU likely cannot deliver rated wattage
Burning smell	Shut down and replace immediately

The Complete PSU Diagnostic Script (Windows)

psu_check.ps1 - Click to expand

# PSU / Voltage Health Check
# Run from an elevated (Administrator) PowerShell session

Write-Host "============================================"
Write-Host "  PSU / VOLTAGE DIAGNOSTICS"
Write-Host "  $(Get-Date)"
Write-Host "============================================"

# 1. Hardware Monitor Sensors (requires LibreHardwareMonitor running)
Write-Host "`n--- VOLTAGE SENSORS ---"
$found = $false
foreach ($ns in @("root\OpenHardwareMonitor", "root\LibreHardwareMonitor")) {
    try {
        $sensors = Get-WmiObject -Namespace $ns -Class Sensor -ErrorAction Stop |
            Where-Object { $_.SensorType -eq "Voltage" }
        if ($sensors) {
            $found = $true
            $sensors | ForEach-Object {
                $val = [math]::Round($_.Value, 3)
                $min = [math]::Round($_.Min, 3)
                $max = [math]::Round($_.Max, 3)
                Write-Host "  $($_.Name): ${val}V (min: ${min}V, max: ${max}V)"
            }
        }
        break
    } catch { continue }
}
if (-not $found) {
    Write-Host "  No hardware monitor found. Run LibreHardwareMonitor for voltage data."
}

# 2. Kernel-Power Event 41 Analysis
Write-Host "`n--- UNEXPECTED SHUTDOWN ANALYSIS (last 30 days) ---"
$kp41 = Get-WinEvent -FilterHashtable @{
    LogName      = 'System'
    ProviderName = 'Microsoft-Windows-Kernel-Power'
    Id           = 41
    StartTime    = (Get-Date).AddDays(-30)
} -ErrorAction SilentlyContinue

$powerLossCount = 0
$bsodCount = 0

if ($kp41) {
    foreach ($ev in $kp41) {
        $xml = [xml]$ev.ToXml()
        $eventData = $xml.Event.EventData.Data
        $bugcheckCode = ($eventData | Where-Object { $_.Name -eq "BugcheckCode" }).'#text'
        $sleepInProgress = ($eventData | Where-Object { $_.Name -eq "SleepInProgress" }).'#text'

        if ($bugcheckCode -eq "0") {
            $powerLossCount++
            $type = "POWER LOSS"
        } else {
            $bsodCount++
            $type = "BSOD (code: $bugcheckCode)"
        }

        $sleepText = if ($sleepInProgress -ne "0") { " [during sleep]" } else { "" }
        Write-Host "  $($ev.TimeCreated) - $type$sleepText"
    }
    Write-Host ""
    Write-Host "  Summary: $powerLossCount power loss, $bsodCount BSOD events"
    if ($powerLossCount -gt $bsodCount) {
        Write-Host "  => Power loss events dominate - investigate PSU and mains power" -ForegroundColor Red
    }
} else {
    Write-Host "  No unexpected shutdowns found."
}

# 3. ATX Voltage Reference
Write-Host "`n--- ATX VOLTAGE REFERENCE ---"
Write-Host "  Rail      Nominal    Tolerance    Danger Zone"
Write-Host "  +3.3V     3.300V     3.135-3.465  below 3.1 / above 3.5"
Write-Host "  +5V       5.000V     4.750-5.250  below 4.7 / above 5.3"
Write-Host "  +12V     12.000V    11.400-12.600 below 11.0 / above 13.0"
Write-Host "  +5VSB     5.000V     4.750-5.250  standby rail"

The Complete PSU Diagnostic Script (Linux)

psu_check.sh - Click to expand

#!/bin/bash
# PSU / Voltage Health Check for Linux
# Usage: sudo ./psu_check.sh

echo "============================================"
echo "  PSU / VOLTAGE DIAGNOSTICS"
echo "  $(date)"
echo "============================================"

# 1. Voltage Sensors
echo ""
echo "--- VOLTAGE SENSORS ---"
if command -v sensors &>/dev/null; then
    sensors | grep -iE 'volt|V$|in[0-9]|VCore|AVCC|3VSB|Vbat|\+3\.3|\+5\.|\+12\.'
else
    echo "  lm-sensors not installed. Run: sudo apt install lm-sensors && sudo sensors-detect"
fi

# 2. ACPI / Power Events
echo ""
echo "--- POWER-RELATED KERNEL MESSAGES ---"
dmesg -T 2>/dev/null | grep -iE 'ACPI.*power|voltage|over.?current|under.?voltage|power.supply' | tail -10
if [ $? -ne 0 ]; then
    echo "  No power-related kernel messages found."
fi

# 3. Boot History (look for unclean shutdowns)
echo ""
echo "--- RECENT BOOTS ---"
journalctl --list-boots 2>/dev/null | tail -10

echo ""
echo "--- UNCLEAN SHUTDOWNS (last 30 days) ---"
power_loss=0
clean_shutdown=0

boot_ids=$(journalctl --list-boots --no-pager 2>/dev/null | awk '{print $2}')
for bid in $boot_ids; do
    last_line=$(journalctl -b "$bid" --no-pager 2>/dev/null | tail -5 | grep -c "Reached target.*Shutdown\|System halted\|Power down")
    if [ "$last_line" -eq 0 ]; then
        power_loss=$((power_loss + 1))
    else
        clean_shutdown=$((clean_shutdown + 1))
    fi
done
echo "  Clean shutdowns: $clean_shutdown"
echo "  Unclean/power loss: $power_loss"

# 4. Reference
echo ""
echo "--- ATX VOLTAGE REFERENCE ---"
echo "  Rail      Nominal    Tolerance        Danger Zone"
echo "  +3.3V     3.300V     3.135-3.465V     below 3.1 / above 3.5"
echo "  +5V       5.000V     4.750-5.250V     below 4.7 / above 5.3"
echo "  +12V     12.000V    11.400-12.600V    below 11.0 / above 13.0"
echo "  +5VSB     5.000V     4.750-5.250V     standby rail"

echo ""
echo "--- RECOMMENDATIONS ---"
echo "  1. Run 'sensors' and compare voltages against the reference table"
echo "  2. Use 'stress-ng' to load the system and watch for voltage drops"
echo "  3. If voltages sag more than 5% under load, consider PSU replacement"
echo "  4. For definitive testing, use a multimeter on Molex/SATA connectors"

Quick Reference

Task	Windows	Linux
Read voltage sensors	HWiNFO64 / LibreHardwareMonitor	`sensors` (lm-sensors)
Continuous monitoring	HWiNFO64 CSV logging	`collectd` / `telegraf` / script loop
Crash analysis	Event 41 BugcheckCode parsing	`journalctl --list-boots` + shutdown analysis
Stress test PSU	OCCT (PSU mode)	`stress-ng` + `gpu-burn`
Measure voltages	Multimeter at Molex/SATA	Multimeter at Molex/SATA
Check standby rail	Event 41 SleepInProgress field	Wake-from-sleep failure logs