You launched a game after pushing your CPU, GPU, or RAM a little harder and now it hard-crashes to desktop, freezes with a buzzing audio loop, or throws a DirectX or driver timeout error within seconds of loading into a match. Sometimes you get a black screen that recovers with a “Default WattMan settings have been restored” or “Application has been blocked from accessing graphics hardware” message. Other times the system fully reboots or shows a Blue Screen with WHEA_UNCORRECTABLE_ERROR or VIDEO_TDR_FAILURE.
This is not a wait-and-see problem. An unstable overclock can corrupt save files, shader caches, and even your Windows install if it keeps writing bad data or crashing mid-update. The good news is most post-overclock crashes are reversible without hardware damage if you step back through your settings methodically. As of 2026, both NVIDIA and AMD drivers will automatically retain a failed overclock profile unless you manually clear it, so follow the steps below in order instead of jumping straight to reinstalling Windows.
Likely Causes
| Cause | How to confirm | Fix difficulty |
|---|---|---|
| Unstable GPU core or memory overclock | Crash only in 3D games, passes desktop use; reverts when you click Reset in MSI Afterburner | Easy |
| Unstable CPU overclock or undervolt | WHEA errors in Event Viewer, crashes across multiple games, fails Cinebench or OCCT CPU test | Easy |
| Unstable RAM / XMP / EXPO profile | Random crashes, different error codes each time; fails TestMem5 or Karhu RAM Test | Moderate |
| Corrupted shader cache or driver state after crash | Game crashes on same loading screen even at stock clocks; other games run fine | Easy |
| Insufficient voltage or power limit throttling | GPU-Z shows PerfCap Reason = VRel or Pwr; crashes under load but not at idle | Moderate |
| Overheating triggering thermal protection | HWiNFO64 shows GPU Hot Spot over 100C or CPU package over 95C right before crash | Easy |
Fix 1: Reset Your Overclock to Stock in Software
What causes it
Most crashes after overclocking come from an overclock utility re-applying the unstable settings at every boot. MSI Afterburner, ASUS GPU Tweak III, EVGA Precision X1, Intel XTU, and AMD Software: Adrenalin Edition all have an “Apply at startup” toggle that will keep pushing the bad clock even after a crash.
How to check
Open your tuning app and look at the current values. In MSI Afterburner, check the Core Clock (MHz) and Memory Clock (MHz) sliders – if either shows +100 or more, it is not stock. In AMD Software go to Performance > Tuning > Custom and see if GPU Tuning is set to Enabled. In Intel XTU look under Advanced Tuning to see if Processor Core Ratio or Core Voltage Offset is not at 0.00V.
What to do
1. Open MSI Afterburner and click the circular Reset arrow button in the bottom center, then click the checkmark to apply. Uncheck the Windows icon for “Apply overclocking at system startup” on the left side.
2. For AMD GPUs, open AMD Software: Adrenalin Edition > Performance > Tuning > click Reset or set Tuning Control to Default.
3. For Intel CPU, open Intel XTU > Profiles > select Default profile > Apply. For AMD CPU, open Ryzen Master > Creator Mode > Reset, or simply restart and clear the setting.
4. Restart your PC and launch the game without re-opening the tuner.
How to undo it
Re-open your tuner and re-apply your last stable profile. In MSI Afterburner click Profile > Save to slot 1 before you reset, so you can click that slot to restore it later. In Ryzen Master your previous values are saved under Profile 1 and Profile 2.
Fix 2: Load Optimized Defaults in BIOS / UEFI
What causes it
Manual BIOS overclocks for CPU multiplier, BCLK, CPU voltage, or memory XMP/EXPO persist outside of Windows. Even if you reset Afterburner, a BIOS-level overclock will keep the system unstable in every game.
How to check
Restart and tap Delete or F2 repeatedly during boot to enter BIOS. On ASUS boards look at the top bar for Ai Overclock Tuner, on MSI look for Overclocking > OC Explore Mode, on Gigabyte look for Tweaker. If CPU Ratio shows anything other than Auto or your CPU’s default (for example 55x on a 13600K or 52x on a 7800X3D) or DRAM Status shows DDR5-6000+ with XMP Enabled, a BIOS overclock is active.
What to do
1. In BIOS press F5 for Load Optimized Defaults or go to Exit > Load Optimized Defaults > Yes.
2. If you need RAM at its rated speed, you can re-enable it after testing: go to Ai Tweaker / OC / Tweaker > Ai Overclock Tuner / DRAM Profile > set to XMP I or EXPO I, leave CPU settings on Auto for now.
3. Press F10 > Yes to Save and Exit.
4. If you cannot reach BIOS due to instant crashes, clear CMOS: power off, flip PSU switch off, hold the case power button for 10 seconds, then either short the CLRTC / JBAT1 pins with a screwdriver for 10 seconds or remove the coin-cell battery for 5 minutes. Power back on.
How to undo it
Enter BIOS again and select your saved OC Profile under Tool > Overclocking Profile or set Ai Overclock Tuner back to Manual and re-enter your previous multiplier and voltage. If you cleared CMOS, your previous profile is erased but you can rebuild it from photos or notes.
Fix 3: Clean Reinstall Your Graphics Driver
What causes it
A crash during a 3D load can leave the driver in a corrupted state. The driver may report successful clocks but fail to compile shaders, causing every subsequent launch to crash even at stock speeds.
How to check
Open Event Viewer > Windows Logs > System and look for nvlddmkm or amdkmdap errors timed to your crashes. In Device Manager > Display adapters, right-click your GPU > Properties > Events tab – if you see “Device not started” after a crash, driver corruption is likely. Also check C:\Users\YourName\AppData\Local\NVIDIA\DXCache or AMD\DxCache for crash timestamps.
What to do
1. Download the latest driver from NVIDIA or AMD but do not install it yet.
2. Disconnect from the internet to prevent Windows Update from installing a generic driver.
3. Boot into Safe Mode: Settings > System > Recovery > Advanced startup > Restart now > Troubleshoot > Advanced options > Startup Settings > Restart > press 4 for Safe Mode.
4. Run Display Driver Uninstaller (DDU) > Select Device Type GPU > Select NVIDIA or AMD > Click Clean and restart.
5. Back in normal Windows, run the driver installer you downloaded, choose Custom (Advanced) > check Perform a clean installation for NVIDIA or Factory Reset for AMD, and install without GeForce Experience overlay if you suspect overlay conflicts.
How to undo it
You cannot undo a clean install, but you can roll back: Device Manager > Display adapters > right-click GPU > Properties > Driver tab > Roll Back Driver. Or re-run the installer for your previous version found in C:\NVIDIA\DisplayDriver or downloaded from the archive.
Fix 4: Clear Shader Cache and Verify Game Files
What causes it
Games pre-compile shaders based on your hardware state. An unstable overclock can write bad shaders to disk. After you return to stock, the game still loads the broken cache and crashes at the same spot.
How to check
The crash is repeatable at a specific loading screen, map, or shader compilation step, while synthetic benchmarks like 3DMark or FurMark now run fine at stock. Check the game’s crash log in Documents\My Games or %AppData% for DXGI_ERROR_DEVICE_REMOVED or shader compilation failure.
What to do
1. Clear driver shader caches: For NVIDIA, open NVIDIA Control Panel > Manage 3D settings > Shader Cache Size > click Clean Shader Cache or manually delete C:\Users\YourName\AppData\Local\NVIDIA\GLCache and DXCache. For AMD, open AMD Software > Settings > Graphics > Reset Shader Cache > Perform Reset.
2. Clear Windows DirectX cache: Settings > System > Storage > Temporary files > check DirectX Shader Cache > Remove files. You can also run Disk Cleanup > Clean up system files > DirectX Shader Cache.
3. Clear the game’s cache: In Steam, right-click game > Properties > Installed Files > Verify integrity of game files. For Battle.net or EA App use Scan and Repair. Delete the game’s local shader folder if listed in support docs, commonly found in C:\Users\YourName\AppData\Local\GameName\Saved or Documents\GameName\shadercache.
4. Relaunch the game and let it recompile shaders without interruption.
How to undo it
Verification and cache clearing cannot be undone, but the game will rebuild all files automatically. If you deleted a custom shader mod, restore it from your Recycle Bin or mod manager.
Fix 5: Dial Back Memory and Fabric Clocks
What causes it
Memory overclocks cause the most confusing crashes because they are not tied to one game’s engine. An XMP / EXPO profile that passed a quick boot test can still fail after 20 minutes of gaming when memory gets hot, producing random error codes like EXCEPTION_ACCESS_VIOLATION.
How to check
Run a dedicated memory test at your current settings. Use TestMem5 with the 1usmus_v3 config for 3 cycles, or OCCT > Memory test for 30 minutes. In HWiNFO64, check Memory Errors (if visible for your platform) or look for Windows Hardware Error Architecture (WHEA) warnings in Event Viewer. If you see errors within minutes, memory is unstable.
What to do
1. Enter BIOS > Tweaker / OC > set DRAM Frequency from Auto/XMP to a lower step, for example from DDR5-6000 to DDR5-5600 or from DDR4-3600 to DDR4-3200.
2. If you manually tightened timings, set Memory Timings to Auto temporarily. For AMD AM5, also set Infinity Fabric Frequency (FCLK) to 2000 MHz or Auto instead of 2033+.
3. Keep DRAM Voltage as specified on your kit’s label (typically 1.35V for DDR4, 1.25V to 1.35V for DDR5) – do not exceed 1.45V for daily use.
4. Save, boot to Windows, and retest the game. If stable, you can try stepping back up by 200 MHz increments and retesting.
How to undo it
Return to BIOS > Tweaker > Enable XMP / EXPO I again, or manually set your previous DRAM Frequency, timings, and voltage. Save with F10.
Fix 6: Fix Power, Voltage, and Thermal Limits
What causes it
A heavy GPU overclock without enough power limit or voltage will crash the moment the card boosts under load. Similarly, if the cooler cannot keep up, the GPU or CPU will hit its thermal limit and force a driver reset.
How to check
Install HWiNFO64 and run Sensors-only mode. Launch the game in windowed mode and watch GPU Temperature, GPU Hotspot Temperature, CPU Package Temperature, and GPU Power. If Hotspot exceeds 105C on AMD or 100C on NVIDIA, or CPU Package exceeds 95C, you are thermal throttling. In GPU-Z > Sensors, watch PerfCap Reason: if it flips to Thrm, Pwr, or VRel right before the crash, you have a limit problem. Check that your power supply cables are fully seated – two separate PCIe 8-pin cables are preferred over a single daisy-chained cable for cards over 225W.
What to do
1. Return core clocks to stock and leave Power Limit at 100% in Afterburner instead of 110-133%. Do not increase Core Voltage beyond +0 mV until stability is confirmed.
2. Set a more aggressive fan curve: MSI Afterburner > Settings (gear icon) > Fan > Enable user defined software automatic fan control > set 60% at 60C, 80% at 75C, 100% at 85C > Apply.
3. Reseat power cables and ensure the GPU is in the top PCIe x16 slot with the retention clip locked. For 12VHPWR / 12V-2×6 cards, verify the connector is fully inserted with no gap and no sharp bends within 35mm of the plug.
4. Improve case airflow: remove dust filters and front panel temporarily to test, verify front fans are intake and rear/top are exhaust.
How to undo it
In Afterburner, restore your custom fan curve by unchecking user defined control or loading your saved profile. Reset Power Limit to your previous value. Reinstall case panels and filters.
Fix 7: Test Stability Before Going Back Up
What causes it
Going straight back into the same game to test an overclock wastes time because games crash inconsistently. A proper stress test finds instability in minutes without risking save corruption.
How to check
After resetting to stock and confirming the game no longer crashes, you need a baseline that proves the system is actually stable now. If a stress test still fails at stock, you have an underlying hardware or cooling issue, not just a bad overclock.
What to do
1. CPU test: Run OCCT > CPU > Extreme > Variable load > 15 minutes, or Cinebench R23 > Multi Core loop for 10 minutes. Watch for errors or immediate reboots.
2. GPU test: Run 3DMark Time Spy Stress Test or Unigine Heaven in windowed mode for 20 minutes. For NVIDIA, use OC Scanner in Afterburner > Test to detect unstable offsets; for AMD use the built-in Stress Test in Performance > Tuning.
3. RAM test: Run TestMem5 or Karhu for at least 30 minutes if you changed memory settings.
4. Only after all three pass without errors, increase your overclock in small steps: +25 MHz on GPU core, +50 MHz on GPU memory, or +1 on CPU multiplier. Test again after each step. Keep notes in Notepad with each stable combination.
How to undo it
Simply stop the stress tests with the Stop button. No changes are made to files, so there is nothing to undo. If you raised clocks again, click Reset in your tuner or reload your last known-stable profile.
If Nothing Works
If the game still crashes at fully stock settings after a DDU reinstall and shader wipe, check Windows for deeper damage. Open Command Prompt as Administrator and run sfc /scannow, then DISM /Online /Cleanup-Image /RestoreHealth, then restart. Check Event Viewer > Windows Logs > Application for .NET Runtime or ntdll.dll errors that point to corrupted system components. Try the game on a different drive or with a new Windows user profile to rule out profile corruption.
Test with a different game or a benchmark at stock. If everything crashes, inspect hardware: reseat RAM sticks in slots A2 and B2, reseat the GPU, and try one RAM stick at a time. Reset CMOS again and run with XMP disabled for a day. If crashes persist only in one title, delete its config folder completely – for Unreal Engine games this is often C:\Users\YourName\AppData\Local\GameName\Saved\Config\WindowsClient – and let it regenerate. As a last resort, use System Restore to a restore point from before you overclocked: Control Panel > Recovery > Open System Restore.
FAQ
Will a failed overclock damage my GPU or CPU permanently?
In almost all cases no, if you revert promptly. Modern GPUs and CPUs have built-in thermal and power protection that shuts the card down before physical damage. Damage risk rises only if you raised voltage beyond safe daily limits and left it there for extended periods, or if temperatures were consistently above 100C. Resetting to stock and fixing cooling is sufficient for most users.
Why does my overclock pass a stress test but still crash in one game?
Stress tests use consistent loads, while games have rapid power and clock spikes during shader compilation, loading screens, and scene transitions. A game with ray tracing or heavy compute shaders can trigger a transient power draw that FurMark does not. Lower your memory clock by 100 to 200 MHz first, then your core boost by 25 MHz, as memory instability shows up most often in specific engines.
Do I need to disable XMP / EXPO completely?
Not always. First test with XMP disabled to confirm the system is stable at JEDEC speeds like DDR5-4800. If the crashes stop, your kit needs manual tuning rather than full auto. Re-enable XMP but lower the frequency one step, set SoC voltage to Auto, and ensure your BIOS is on the latest stable version for improved memory compatibility.
Should I use Auto overclocking features like PBO, Boost, or NVIDIA OC Scanner?
Leave them off until you have confirmed stability at stock. AMD Precision Boost Overdrive, Intel Thermal Velocity Boost, and NVIDIA’s automatic OC scanner each add their own offsets on top of manual settings and can stack to an unstable total. Once stable, enable only one at a time. In Ryzen Master set PBO to Disabled, and in Afterburner avoid clicking OC Scanner until manual offsets are stable.
