PC Restarts Only During Games Find the Faulty Part with OCCT
Your PC restarts only inside games and never shows an error. That pattern has a short list of causes, and OCCT can test each one separately - GPU heat, graphics memory, CPU, and finally the power supply under combined load.
✍️ Technical Anshu🧩 Windows 10 / 11⏱️ 10 minute read
🎥 Watch the full video - every OCCT test shown live, step by step
The PC is perfect on the desktop, perfect in videos, perfect in light apps - then a game starts, the fans spin up, and the whole machine restarts without any error. This pattern almost always means one part fails only at full gaming load: the graphics card, the power supply feeding it, or the heat that builds up around both. Guessing and replacing parts randomly is expensive. Instead, reproduce the failure in a controlled way with OCCT, one test at a time, in this order.
Gaming load vs stress test - example readings at the moment of failure
GPU temp89 C
GPU hotspot104 C
+12V rail11.4 V
CPU temp67 C
System stateRESTART
Example readings just before a restart. A very hot GPU hotspot or a +12V rail dropping well under 12V is the clue to follow.
1
First Rule Out Heat - GPU 3D Test for 20 Minutes
Start with the GPU 3D test in OCCT and watch two numbers: GPU temperature and hotspot temperature. The hotspot is the hottest spot on the chip and can run 15-25 degrees above the main reading.
Run the GPU 3D test and note both temperatures every few minutes.
If the PC restarts here, note the exact temperature at that moment.
Restarting at roughly the same high temperature each time means heat - clean the card, check its fans actually spin, and check case airflow before blaming anything else.
If it survives 20 minutes at reasonable temperatures, heat is unlikely to be the whole story. Continue.
2
Test the Graphics Memory - VRAM Test
Games fill the graphics memory far more than desktop apps do, and faulty VRAM often only complains under a full, sustained load. Run OCCT's VRAM test at a high memory usage setting for 20 to 30 minutes.
Any error count above zero here is a verdict: the GPU's memory is faulty. No driver reinstall fixes bad memory chips. If the card is in warranty, this result is your evidence.
3
Load the CPU Alone - CPU Test for 30 Minutes
Games also load the CPU hard, and a restart blamed on the GPU is sometimes a CPU or motherboard power stage giving up. Run the CPU test (OCCT mode, Large data set) for 30 minutes.
If the CPU test alone restarts the PC, the graphics card is innocent. Look at CPU temperature first, then the motherboard's power delivery, then the PSU. If CPU errors or WHEA warnings appear instead of a restart, the CPU is unstable at its current voltage - reset any overclock or undervolt and retest.
4
The Final Exam - Power Test, Short and Supervised
If each part passed alone, run the Power test - CPU and GPU together - for 10 to 15 minutes only, sitting with the PC. This is the closest match to a game's real demand on the power supply.
Watch the +12V rail graph. A healthy line stays close to 12V under load.
A restart that happens only in this combined test, with a sagging 12V line, points squarely at the PSU - it handles each part alone but collapses when both draw together.
Stop the test at once if you smell anything electrical or the rail drops sharply.
Never run this test repeatedly or for long on a PSU you already suspect - one clear result is all the evidence you need.
🔴 Read this before any stress test: a stress test pushes hardware harder than any game. Stay with the PC while it runs, keep the temperature graph on screen, and stop the test at once if you smell burning, hear crackling, or see temperatures racing past 100°C. Always test at stock settings first (no overclock), and never run a long Power test on a power supply you already suspect is faulty — on a suspect PSU, a short supervised run is the most you should attempt.
5
Reading the Outcome
Restarts in GPU test at high temp - cooling first: dust, fan curve, dried thermal paste on the card.
VRAM errors - faulty graphics memory; warranty or replacement, not settings.
Restarts or errors in CPU test - CPU cooling, voltage settings, or motherboard power stage.
Restarts only in Power test - PSU is the prime suspect; confirm by testing with a known-good PSU if possible.
Everything passes - the fault may be a game/driver issue or a rare event; test longer and check Windows Event Viewer for WHEA entries after real gaming sessions.
🎥 Watch the Complete Video for Full Knowledge and Installation
Every test in this guide is shown live in the video - download, setup, CPU, GPU, RAM and PSU testing with real readings.
Why does my PC restart only in games and not in stress tests on the desktop?
Games are the heaviest real load most PCs ever face - GPU, CPU and power draw peak together. A weak PSU, hot GPU hotspot or faulty VRAM can survive light work for years and only fail in that combined peak, which is exactly what OCCT reproduces one part at a time.
There is no blue screen, just a restart. What does that mean?
A restart with no blue screen and no error usually means power or protection: the PSU gave up, or a temperature/short protection cut power instantly. Windows never got the chance to complain. That is why the Power test and the temperature readings matter more than error logs for this symptom.
OCCT passed everything but games still restart the PC. Now what?
Run each test for the full free hour, since rare faults need time. Then check Event Viewer for WHEA errors logged during real games, test the RAM with OCCT's Memory test, and try a known-good PSU - a PSU that passes an hour can still be the weakest link, so swapping is the final confirmation.
This guide is based on the Technical Anshu video on testing a PC with OCCT • Stuck on any step? Ask in the video comments - you will get an answer there 🙏