Why Your PC Shuts Down Randomly: Thermal Causes

A PC that shuts down by itself is not malfunctioning — it is protecting itself. Modern hardware has built-in thermal protection that triggers an emergency shutdown when temperatures exceed safe limits. Understanding where those limits are, which component crosses them first, and how to read the sensor data afterward is the difference between guessing and diagnosing. This guide is part of our complete PC temperature reference — but focused specifically on the scenario where the machine shuts off entirely rather than throttles.

After 8 years of diagnosing machines in Copenhagen, random shutdowns that are genuinely random — not thermal, not PSU, not driver — account for less than 10% of the cases we see. The other 90% have a measurable cause in sensor data, even if the sensor data was never recorded. Here is a structured investigation framework, starting with the most common causes and working toward the rare ones.

Why PCs Shut Down Versus Why They Throttle

Thermal throttling and thermal shutdown are two different protection mechanisms. It helps to understand both.

Thermal throttling begins before the danger zone: a CPU running at 90-95°C reduces its clock speed to shed heat. The machine stays on, but runs slower. The user might notice sluggishness, but work continues. Our complete guide to thermal throttling covers this mechanism in detail.

Thermal shutdown is the emergency brake. When a component exceeds its absolute maximum temperature despite throttling, the system cuts power completely — no graceful shutdown, no warning dialog. Just off. This protects the hardware from physical damage that would otherwise be permanent.

Component	Throttle starts	Emergency shutdown
Intel Core i9-14900K	~95°C	100°C (TjMax)
Intel Core Ultra 285K (Arrow Lake)	~100°C	105°C (TjMax)
AMD Ryzen 9 9950X	~90°C	95°C (TjMax)
AMD Ryzen 7800X3D (X3D)	~84°C	89°C
NVIDIA RTX 4090 (core)	83°C	~95-100°C
NVIDIA RTX 4090 (hotspot)	~100°C	~110°C
NVMe SSD (typical Gen 4)	~70°C	~80°C

The machine that shuts down at 95°C every time it runs a game has a CPU hitting TjMax. The machine that shuts down after 20 minutes of idle may have a PSU failing under sustained load — a completely different cause, same symptom.

The Most Common Causes, in Order

In our monitoring data across hundreds of machines, here is how random shutdowns distribute by root cause:

Cause	Estimated frequency	Detectable by sensors
CPU thermal shutdown	35-40%	Yes — CPU temp at shutdown
PSU failure under load	20-25%	Partially — voltage rails
GPU thermal shutdown	15-20%	Yes — GPU temp at shutdown
Fan failure (any)	10-15%	Yes — fan RPM
VRM thermal shutdown	5-10%	Yes — VRM temp
RAM failure	3-5%	No — requires memory test
Software/driver cause	<5%	Partially — event log

1. CPU Thermal Shutdown

This is the single most common cause. The CPU reaches TjMax and the system cuts power before physical damage occurs. The diagnostic is straightforward: if a machine consistently shuts down after 10-30 minutes of sustained CPU load (video encoding, compilation, gaming, large spreadsheets), and the CPU was approaching its TjMax before shutdown, the cause is thermal.

Why the CPU reaches TjMax:

• Dried thermal paste — the most common cause in machines over 2-3 years old. Thermal compound degrades from gel to powder, losing 60-80% of its conductivity. The CPU temperature under load rises by 15-30°C compared to a fresh application.
• Clogged heatsink fins — dust acts as insulation. A heatsink with dust-packed fins moves less heat even with a perfectly functional fan. Our overview of all monitored sensor types explains why CPU package temperature is one of the seven sensors you should never leave unwatched.
• Inadequate cooling for the CPU's TDP — a 125W CPU on a 65W cooler will hit TjMax under sustained all-core load regardless of dust or paste condition.
• Case airflow failure — fans removed, vents blocked, or a case designed for poor airflow.

How to confirm it: Check the Windows Event Log immediately after a shutdown. Event ID 41 (Kernel-Power) appears for unplanned shutdowns. If the CPU was running near TjMax before the shutdown — visible in HWiNFO64 logs if enabled, or in GGFix telemetry if the machine is monitored — the cause is confirmed.

2. PSU Failure Under Load

A power supply that measures fine at idle can fail to deliver stable voltage under load. When a PSU cannot maintain the 12V, 5V, or 3.3V rails within their ATX specification tolerances (+/- 5% for 12V), the system triggers a brown-out protection shutdown.

This is harder to diagnose than thermal shutdown because the failure is intermittent and load-dependent. The pattern: machine runs fine for light tasks, shuts down during gaming, video editing, or any sustained high-CPU + high-GPU workload. No thermal event in the logs. CPU and GPU temperatures are normal.

PSU failure modes that cause random shutdowns:

• Aging capacitors losing their ability to filter voltage ripple
• PSU operating near or at its rated maximum wattage continuously
• High ambient temperatures reducing the PSU's effective output capacity
• A partially failed secondary rail that only drops out under specific load combinations

How to confirm it: Measure the 12V rail directly under sustained load with a multimeter, or use hardware monitoring to watch the 12V reading in BIOS/HWiNFO. A reading below 11.4V (+5% below nominal) under load is a failing PSU. Alternatively: swap the PSU for a known-good unit and test. If the shutdowns stop, the PSU was the cause.

3. GPU Thermal Shutdown

The GPU has its own thermal protection independent of the CPU. An RTX 4090 running at 92°C core temperature is within tolerance. The same GPU with a blocked exhaust, failed fan, or degraded thermal interface running at 98°C+ will trigger an emergency shutdown.

GPU shutdowns have a distinctive pattern: they occur during graphically intensive workloads — gaming, 3D rendering, GPU compute — but not during CPU-only tasks. The GPU temperature at the time of shutdown (visible in post-mortem Windows event logs or continuous monitoring data) confirms the cause.

Common GPU thermal shutdown triggers:

• Fan failure — a single failed GPU fan reduces cooling capacity by 30-50% depending on design
• GPU hotspot exceeding limits — even when the edge temperature looks acceptable at 83°C, the hotspot runs ~20°C higher. A hotspot at 105°C+ triggers protection on most cards.
• Thermal paste degradation on the GPU die (relevant for cards over 4 years old)
• Restricted airflow inside the case, particularly for blower-style GPU coolers

4. Fan Failure

A fan that fails does not always stop immediately and visibly. Bearing failure is progressive: the fan runs more slowly, draws more current, produces more noise, and eventually seizes. In the interim weeks or months, reduced airflow causes rising temperatures across all components simultaneously.

In our monitoring data, a CPU fan dropping from its rated 1,800 RPM to 900 RPM correlates with a 12-18°C rise in CPU temperatures under identical workload conditions. If the fan seizes entirely, the temperature rise is immediate and severe — often triggering shutdown within 5-15 minutes of any sustained workload.

Fan failure is one of the cleanest cases for predictive monitoring. The RPM decline is gradual and clearly visible in sensor trends weeks before failure. A fan running at 60% of rated speed is a scheduled maintenance item. A fan that has already seized is an emergency.

5. VRM Thermal Shutdown

The voltage regulator modules on the motherboard convert the 12V input to the precise voltage the CPU requires — typically 0.8-1.4V depending on load state. VRMs generate significant heat in the process. When they overheat, the motherboard's protection circuit triggers an emergency shutdown to prevent permanent damage.

VRM shutdown is less common than CPU or GPU thermal shutdown but follows a similar pattern: it occurs under sustained high-CPU load on machines where the VRM area has poor airflow, damaged or missing heatsinks, or where the CPU is drawing more power than the VRM was designed to handle (overclocking scenarios, or pairing a high-TDP CPU with a budget motherboard).

Our detailed VRM temperature guide covers the specific temperature thresholds and how to identify VRM-related shutdowns in sensor data.

The Diagnostic Process: Step by Step

When a machine reports random shutdowns, this is the investigation sequence that reliably identifies the cause:

Step 1: Check the Windows Event Log Open Event Viewer → Windows Logs → System. Look for Event ID 41 (Kernel-Power, "The system has rebooted without cleanly shutting down first") and Event ID 1001 (BugCheck, if a BSOD occurred). Note the timestamp and any preceding events in the 60 seconds before the shutdown.

Step 2: Enable hardware sensor logging Before the next shutdown, install HWiNFO64 and enable sensor logging at 1-second intervals. After the next shutdown, open the log and look at the 30-60 seconds before the shutdown timestamp. Which component was closest to its maximum? That is your suspect.

Step 3: Reproduce under controlled load Run a stress test — Prime95 for CPU, FurMark for GPU — while watching temperatures in real time. Does the machine shut down during CPU stress (CPU thermal), GPU stress (GPU thermal), or both simultaneously (PSU under combined load)?

Step 4: Inspect physically Before ordering parts: open the case. Is every fan spinning? Is the heatsink clogged with dust? Is the GPU fan running? Are all power connectors fully seated? Visual inspection eliminates a surprising number of cases.

Step 5: Confirm the fix After the repair — thermal paste replacement, fan replacement, PSU swap, dust cleaning — run the same stress test that previously caused the shutdown. Monitor temperatures throughout. If the machine completes a 30-minute stress test without shutting down and temperatures stay below throttle thresholds, the fix is confirmed.

For fleet environments, continuous monitoring eliminates steps 1-2 by maintaining historical sensor data automatically. When a client calls reporting a shutdown, GGFix logs already contain the temperature readings from the minutes before the event — no guessing, no waiting for the next occurrence to instrument the machine.

When Shutdowns Are Not Thermal

Not every random shutdown is thermal. If you have confirmed that temperatures are normal throughout the failure scenario, investigate these:

RAM instability — Bit errors in RAM cause kernel panics that result in immediate shutdown. Unlike thermal events, these occur at any load level, often during memory-intensive operations like opening many browser tabs or loading large files. Run MemTest86 for 2+ passes to confirm or rule out.

Driver-triggered BSOD — A driver crash in kernel mode causes an immediate system reset. The difference from a thermal shutdown: a BSOD shutdown triggers a memory dump (visible in Event Viewer as BugCheck). Check the stop code — GPU driver crashes are common and identified by codes referencing display adapter failures.

Corrupt Windows system files — Run sfc /scannow in an elevated command prompt to check system file integrity. Rarely the cause of shutdowns but worth ruling out after hardware checks are clean.

For a complete framework of all slow-PC and shutdown causes, our troubleshooting guide covers both the hardware and software diagnostic approaches end-to-end.

How Continuous Monitoring Prevents Recurrence

A thermal shutdown is recoverable — the protection mechanism worked, no hardware was damaged. But the machine that shuts down today will shut down again tomorrow under the same conditions. Worse, repeated thermal shutdowns indicate sustained high-temperature operation that accelerates component aging even when the shutdown fires before physical damage.

The pattern we see in monitored fleets: machines that have ever triggered a thermal shutdown almost always show the warning signs for 2-4 weeks before the first event — a CPU idle temperature that has risen from 45°C to 68°C over months, a fan RPM that has declined from 1,400 to 900 RPM over the same period, a CPU temperature under load that has trended from 82°C to 96°C.

None of those trends are visible without continuous logging. Each individual snapshot looks within spec. Only the trend reveals the trajectory toward failure.

GGFix monitors CPU temperature, GPU temperature, fan RPM, VRM temperature, and NVMe temperature continuously, flags anomalous trends automatically using AI-based pattern recognition, and sends alerts via Telegram, Slack, or email before temperatures approach shutdown thresholds — not after the machine has already gone dark.

Frequently Asked Questions

Q: How do I find out what caused my PC to shut down?

Open Event Viewer (search for it in the Start menu) → Windows Logs → System. Look for Event ID 41 labeled "Kernel-Power" with a "BugcheckCode" of 0. This confirms an unexpected power loss. The timestamp tells you exactly when it happened. Cross-reference with CPU/GPU temperature logs (if you have them enabled in HWiNFO64 or similar) to identify which component was near its limit.

Q: At what temperature does a CPU shut down automatically?

Most modern Intel desktop CPUs (12th-14th gen) have a TjMax of 100°C. Intel Core Ultra 200S (Arrow Lake) is 105°C. AMD Ryzen 7000/9000 series is 95°C, with the X3D variant at 89°C. The CPU begins throttling 5-10°C before TjMax and triggers emergency shutdown when the limit is reached despite throttling.

Q: Can a failing PSU cause random shutdowns without any thermal event?

Yes — and this is one of the most misdiagnosed shutdown causes. A PSU that cannot maintain stable voltage on the 12V rail under high GPU and CPU load triggers a brown-out shutdown that looks identical to a thermal event but leaves no temperature spike in the logs. The diagnostic is to monitor voltage rails under sustained load — a 12V reading below 11.4V is outside the ATX +/-5% tolerance and indicates a failing PSU.

Q: My PC only shuts down when gaming. What does that tell me?

Gaming loads the GPU to near-maximum simultaneously with significant CPU demand. The combination draws peak power from the PSU while generating peak heat from both components. This pattern points to three suspects in roughly equal likelihood: GPU thermal shutdown, CPU thermal shutdown under combined load, or PSU voltage instability under peak combined power draw. Start by checking GPU temperature at the moment of shutdown.

Q: How often does dust cause random shutdowns?

In our repair data, dust-related thermal issues account for roughly 25-30% of shutdown cases in PCs over 2 years old. The dust itself does not cause the shutdown — it reduces airflow enough that the CPU, GPU, or VRM reaches its shutdown threshold under workloads it previously handled without issue. Cleaning (compressed air, careful disassembly) combined with thermal paste replacement resolves the majority of dust-related shutdown cases permanently.

Q: Can I prevent random shutdowns from recurring after fixing the immediate cause?

Fixing the immediate cause stops the current shutdowns, but the underlying condition — aging thermal paste, degrading fans, a PSU approaching end of life — will return. Continuous hardware monitoring is the only mechanism that catches these trends before they cause the next failure. Monitoring CPU and GPU temperatures, fan RPM, and (for newer machines) NVMe temperature over time converts the next "random" shutdown into a scheduled maintenance visit.