Only One Game Crashes While Others Run Fine? Shader Cache, Game Files & Per-Game Settings Diagnosis Guide

| HK GPU Fix

Only One Game Crashes While Others Run Fine? Shader Cache, Game Files & Per-Game Settings Diagnosis Guide

Key Takeaways

When only one game crashes, blacks out, or shows corrupted graphics while other games run normally on the same machine, card, and driver, that single-game pattern is itself the diagnosis: shader-cache, game-file, per-game-setting, overlay, and save-file tests, run in order, separate a problem that follows the game from a hardware or driver problem that would affect more than one game.

What This Problem Is, and What It Is Not

The symptom is narrow and that narrowness matters. On one machine, with one graphics card and one driver version, a single game crashes back to the desktop, blacks out and returns to the desktop, blacks out and needs to be force-closed, or shows corrupted graphics while still running. Other games, especially games with a similar load, play normally. The problem may strike every time a particular scene loads, or only after a period of play, but its range stays locked to one game.

That range is the most important diagnostic signal in the article. Graphics card hardware faults in the core, memory, power delivery, or cooling do not usually behave politely and target one game only. A fault locked to one game points first to that game side of the system: its files, its settings, its shader cache, its rendering path, or software injected into it, rather than to the hardware. Starting from hardware in this situation wastes effort and often ends with a healthy card being blamed.

The symptom range also needs to be drawn wide enough to be useful. Crashing to the desktop, blacking out and recovering, blacking out and needing a forced close, corrupted graphics with the program still alive, and reaching the main menu but failing on entering gameplay all belong to the same single-game spectrum, and the diagnostic logic in this article covers all of them. One boundary must be stated plainly: if launching the game restarts or powers off the whole machine, that is not this problem. A full-machine restart or power loss points toward power supply and system-level investigation first, and the reader should leave this article for that guidance rather than continuing through game settings.

This article also has neighbours it must not rewrite. A guide on choosing between rendering interfaces for the best smoothness is a choice guide about architecture and performance trade-offs, while this article is a fault-isolation guide in which the interface is only one comparison variable, not the subject. Older games have their own compatibility layer involving old interfaces, 32-bit builds, display scaling, and exclusive fullscreen behaviour, and if the failing game is a title from many years ago, that older-game guidance is the first diversion. A system-wide driver crash that affects everything is a different class from a crash confined to one game, so system-wide driver-crash steps cannot simply be applied here without first proving whether the problem follows the game or follows the driver. A cross-game bottleneck question is also different: if the card were bottlenecked, other games at a similar load should show symptoms too. Video playback decoding is a separate path from three-dimensional game rendering, and browser or office application black screens after enabling hardware acceleration belong to application-layer guidance, even though the overlay on-off comparison method is shared. For completeness, an earlier article on power interruption (#188) contains a single sentence noting that a single-game crash is more likely to involve game-file damage, incomplete driver files, or game settings; that sentence was only a diversion with no diagnostic steps, and this article expands it into a full guide without duplicating that earlier article.

Why Only One Game Can Fail

Six mechanisms explain why a fault can stay locked to one game while everything else works. Each one maps to a test later in the article, so the mechanism section is not background colour; it is the reason the test order exists.

First, the shader and pipeline cache. Modern games compile shaders when a scene is first run, and store the compiled results in a cache. If that cache is damaged, if it becomes invalid after a driver update, or if it no longer matches the game version after a game update, a particular scene can crash on load while other games, which use their own caches, remain unaffected. Clearing the cache is cheap and high-yield, which is why it comes early.

Second, game-file damage. An interrupted download or update, a disk error, or a cloud-sync conflict can damage individual files. File verification can repair those files without reinstalling the whole game, which saves time and download volume. The order here matters and is stated explicitly below: verify first, reinstall only if verification does not resolve the problem.

Third, individual game settings. One rendering interface path may be faulty while another is fine, one upscaling or frame-generation option may trigger the fault, one quality combination may expose it, or exclusive fullscreen versus borderless windowed mode may change the result. Restoring defaults and then adding options back one at a time is the standard way to narrow a trigger condition, because it changes one variable at a time instead of guessing.

Fourth, third-party injection. Overlay software, recording and streaming software, monitoring software, modifications, and anti-cheat interaction with the driver are all variables that follow the game. Other games without those modifications or overlays active naturally stay healthy. That is why an overlay or modification can look exactly like a game bug until it is switched off as a control.

Fifth, save-file or configuration-file damage. An individual save, a configuration file, or an older setting returned by cloud sync can make the game crash every time the same progress point loads. The control is a new save or a default configuration, tried only after backing up the existing files. The reader must never be told to delete saves; backup comes first, always.

Sixth, driver-version interaction with one game. A particular driver version can conflict with a particular game version while other games remain fine. This possibility belongs late in the order, because changing driver versions is costly and because changing the driver before the game-side controls are done can mask the real cause. A driver change made too early destroys the evidence that would have pointed to the game.

The Diagnostic Order: Nine Steps, in This Order

Step 1: Establish That It Really Is Only One Game

Play two or three other games of similar load, using different rendering paths where possible, each for a meaningful session, and record which games fail and which do not. If a second game also fails, the problem is no longer a single-game problem, and the reader needs cross-game or system-level guidance instead. This step is the gate for the whole article: without it, every later conclusion is built on an unproven premise.

Step 2: Record the Trigger Conditions

Before changing anything, record the scene, mode, map, or menu involved, how long into play the fault appears, which graphics settings are active, and which overlays or recording tools are running. The trigger conditions alone narrow the range by half. A fault that appears only with one lighting option enabled is a different trail from a fault that appears on the main menu, and a fault tied to one map points toward that map resources rather than toward the whole game.

Step 3: Clear the Shader Cache

Clear the game shader cache and the driver-level shader cache, then launch the game and let it recompile. If the fault clears, the cause was the cache. Two cautions belong here. First, stutter during recompilation is normal: the game is rebuilding shaders it has just discarded, and that temporary stutter must not be misread as the fault persisting. Second, if the fault returns after the first recompiled run, move on rather than clearing the cache repeatedly; repeated clearing of a cache that is not the cause proves nothing further.

Step 4: Verify the Game Files

Use the platform file-verification function and let it repair damaged files, then test again before considering anything more drastic. Do not skip this step and jump to a full reinstall. Verification repairs individual damaged files in place, while a reinstall costs far more time and download volume for the same result in most damaged-file cases. Only if verification completes and the fault persists should a full reinstall enter the picture.

Step 5: Restore Default Settings, Then Add Options Back One at a Time

Restore the game graphics settings to defaults, paying particular attention to the rendering interface path, upscaling, frame generation, and fullscreen mode, and test with defaults. If the fault persists at defaults, settings are not the trigger, and the investigation moves on. If defaults are healthy, add the previous options back one at a time, testing after each, until the triggering option is found. Adding several options back at once destroys the isolation this step exists to create.

Step 6: Switch Off Third-Party Injection as a Control

Close all overlay, recording, and monitoring software, disable modifications, and test again. If the game becomes healthy, re-enable those items one at a time to find which injection conflicts. The same one-at-a-time rule applies in reverse: switching everything off and then everything on again in one move cannot identify the culprit. Note that modifications must be fully disabled during all other controls as well; any conclusion drawn while modifications are active is unreliable, because the modification itself remains an uncontrolled variable.

Step 7: Try a New Save and Default Configuration, After Backup

Back up the save files and configuration files first, then try a new game progress point or a default configuration. If the new progress point is healthy while the old progress point fails at the same place every time, the direction points to save or configuration damage, not to the graphics card. Cloud-synced settings deserve attention here: a setting synced from another machine can carry the problem with it, so the control is a fresh local configuration, not repeated syncing back and forth between machines.

Step 8: Consider the Driver Version, Last

Only when all game-side controls are complete and the fault still follows this one game should driver version be considered. Check the game official notes and the driver release notes for known issues with this game version, and before trying another driver version, record the current version so it can be restored. A clean driver removal is part of a proper version comparison, and that procedure is covered in system-wide driver-crash guidance, which this article refers to rather than rewriting. The key point is sequence: a driver change is a comparison made after game-side causes are excluded, not a first resort.

Step 9: The Boundary Where This Article Ends

If every game-side control is complete, the fault still persists, and it begins to spread to a second game, or it is joined by system-wide symptoms such as a blue screen, a full-machine restart, or failures in other programs, it is no longer a single-game problem. Continuing to circle through game settings at that point wastes time. The reader needs cross-game and hardware investigation instead, and the record built in steps 1 and 2 is what makes that next investigation efficient.

StepIf the result is healthyIf the fault persists
Play other games of similar loadConfirms the single-game pattern; continue hereSecond game fails too: leave this article for cross-game guidance
Clear the shader cacheCause was cache damage or version mismatchCache is not the cause; move to file verification
Verify game filesCause was damaged filesFiles are sound; move to settings
Restore default settingsA setting is the trigger; add options back one at a timeSettings are not the trigger; move to injection controls
Switch off overlays and modificationsAn injection conflicts; re-enable one at a timeInjection is not the cause; move to save and configuration controls
New save and default configuration after backupSave or configuration damage is the directionSave is not the cause; move to driver-version comparison
Driver-version comparison, lastA driver and game version interaction is the directionGame-side and driver comparisons exhausted: boundary reached

Edge Cases Where the Conclusion Changes

A fault confined to multiplayer mode, one map, or one game mode points to resources specific to that mode: map files, server-synced content, or anti-cheat interaction. The diagnostic method is unchanged, but file verification and settings restoration should be directed at that mode rather than applied blindly to the whole game.

A fault that starts immediately after a game update makes the update itself the prime suspect. The game version, the shader cache version, and the configuration file format may all have changed together. The controls are cache clearing and file verification, and the driver should not be blamed first when the driver has not changed at all.

A fault that starts after a driver update, but still affects only one game, is the combination most easily misread as the driver damaging the card. The correct reading is an interaction between that driver version and that game version, handled by the step 8 version comparison, not by replacing hardware.

Cloud-synced settings and saves need a specific caution. A configuration synced from another machine can import the problem into this machine, and syncing back and forth between machines during testing only spreads the suspect variable. The control is a fresh local configuration on this machine, tested before any further sync.

Modification players have one non-negotiable rule: disable all modifications, then add them back one at a time. No other control in this article is valid while modifications are active, because a modification changes game files and the rendering path itself. Testing with modifications on and concluding that the game is at fault, or that the card is at fault, is not a conclusion at all.

Common Misconceptions, With Corrections

One game crashes, so the graphics card must be failing. That is wrong. Hardware faults do not politely target one game. A fault locked to one game points first to the game side, and the tests in this article exist to prove that direction before hardware is considered.

Reinstalling the driver fixes all game problems. That is wrong. The driver is a shared layer: if it were broken, other games should show symptoms too. Reinstalling the driver before the game-side controls are done wastes time and can mask the real cause, leaving the reader with a fresh driver and the same fault.

File verification is useless; the whole game must be deleted and reinstalled. That is wrong. Verification already repairs individual damaged files, and a full reinstall is the next step only after verification has failed to resolve the problem. Starting with a reinstall spends download volume and time for no diagnostic gain.

The newest driver is always the best choice for every game. That is wrong. Individual game versions and individual driver versions can conflict, so newest does not mean best suited to this particular game. Version suitability is a per-game question, answered by release notes and by the step 8 comparison, not by assuming that newer is universally better.

Clearing the shader cache will damage the game. That is wrong. The cache is regenerable temporary data: clearing it only makes the next load slower while shaders recompile, and it does not damage the game itself. The caution is the opposite one, that recompilation stutter must not be mistaken for the fault surviving.

Testing with overlays and modifications active is fine and does not affect the conclusion. That is wrong. Third-party injection is itself a variable, and any control run with it active cannot separate the game from the injection. Switch injections off for the control, then re-enable them one at a time.

Stop Points and What to Record Before Seeking Inspection

Stop game-side testing if the fault spreads to other games, if the whole machine restarts or powers off, or if a blue screen appears. Those are system-level symptoms, and further game-setting changes will not address them. Also stop before deleting any save or configuration file without a backup: backup is a precondition of step 7, not an optional extra.

Before seeking inspection, record the failing game and its version, the driver version, the other games tested and their results, the exact trigger conditions, the shader-cache result, the file-verification result, the default-settings result and any triggering option found, the overlay and modification results, and the new-save comparison result. That record lets an inspection start from the narrowed field instead of repeating the whole sequence.

For inspection in Hong Kong, the shop is Room 1209, Dali Plaza, Shek Mun. The telephone number is 2186 8010 and the WhatsApp number is 6558 6806.

Frequently Asked Questions

1. Only one game crashes. Does that mean my graphics card is dying?

Very probably not. A hardware fault would not normally confine itself to one game while other games at a similar load run fine. The single-game pattern points first to that game files, cache, settings, injections, or save data, and the ordered tests in this article prove that direction step by step.

2. Should I reinstall the driver first?

No. The driver is shared by all games, so a broken driver should affect more than one game. Do the game-side controls first: cache clearing, file verification, default settings, injection controls, and the save comparison. Driver version is the last comparison, made only when the fault still follows this one game.

3. The game stutters badly the first time I launch it after clearing the shader cache. Is the fault still there?

Not necessarily. That stutter is recompilation: the game is rebuilding the shaders that were just cleared. Judge the result after that first recompiled run has settled, not during it, and do not mistake normal recompilation stutter for a persisting fault.

4. File verification found nothing. Should I reinstall the whole game now?

Not yet. Verification is only one step. Continue through default settings, injection controls, and the save comparison before reinstalling, because a settings trigger, an overlay conflict, or a damaged save will survive a reinstall unchanged or return through cloud sync.

5. The game works at default settings but crashes with my usual settings. How do I find the trigger?

Add the previous options back one at a time, testing after each, starting with the rendering interface path, upscaling, frame generation, and fullscreen mode. The option whose return brings the fault back is the trigger condition, and that single-variable result is far more useful than a vague finding that some setting is involved.

6. The fault only appears in one map or one multiplayer mode. Is that still a single-game problem?

Yes, and it narrows the field further. It points toward resources specific to that mode, such as map files, server-synced content, or anti-cheat interaction. Apply the same sequence, file verification and settings restoration, directed at that mode.

7. The fault started right after a game update. Is the driver to blame?

Unlikely as a first suspect, because the driver has not changed. A game update can change the game version, the shader cache version, and the configuration format together. Clear the cache and verify the files first, before considering any driver factor.

8. The fault started after a driver update, but only one game is affected. Has the driver damaged my card?

No such conclusion follows. The correct reading is an interaction between that driver version and that game version. Handle it with the step 8 version comparison, checking release notes and recording the current version before trying another, rather than replacing hardware.

9. I play with modifications. Can I leave them on while I run the other tests?

No. Modifications change game files and the rendering path, so every other control is invalid while they are active. Disable all modifications first, run the controls clean, and then add modifications back one at a time to find any conflict.

10. I have done every game-side test and the fault is now appearing in a second game. What does that mean?

It means the problem has left the single-game category. Stop circling through game settings and move to cross-game and hardware investigation, taking the full test record with you, including which games fail, the trigger conditions, and every control result so far.

Related articles

Related services

If you need Chip-level GPU repair, see these service pages: