GPU artifacts on screen? Find where the corruption starts

Start by describing what you actually see

“Artifacts” is a broad label. Write down the visible pattern instead: colored squares, flashing dots, broken text, horizontal noise, stretched triangles, missing textures, a frozen frame, or a solid block. Note whether it moves, disappears after a restart, or returns in the same place.

Capture a photo or short video if you can do so without keeping an unstable computer under load. Record the time, what was open, and what changed just before the problem began. A driver update, new game, cable move, dock change, overclock, drop, repair, or cooling problem can change the diagnostic path.

If the symptom is ordinary flicker rather than corrupted shapes or textures, the laptop screen flickering guide has a more focused app, driver, panel, and display-cable workflow.

Where the pattern appears matters more than its shape

It appears before Windows starts

Corruption on the manufacturer logo, firmware screen, or model-supported preboot diagnostic makes one Windows app less likely. It does not automatically identify the graphics card. The monitor, cable, port, integrated graphics, dedicated graphics, memory, power, and broader computer still share the path.

Do not enter firmware menus just to experiment. If the computer already offers a supported preboot or display test, record the result and exit without changing settings.

It starts only after Windows loads

A Windows-only pattern keeps the display driver, operating system, startup software, app overlays, refresh settings, and recent updates in scope. Record whether the sign-in screen is clean and exactly when corruption begins.

Microsoft’s current display guidance uses Task Manager as one clue for flicker and scrambling: if Task Manager changes with the rest of the screen, the display-driver path stays in focus; if Task Manager remains stable, an incompatible app becomes more likely. That comparison is useful evidence, not a universal verdict for every artifact.

It affects one game or app

Test the same scene twice and compare another ordinary app. One repeatable game, browser, editor, or video path can point toward that program, its graphics setting, an extension or overlay, or a driver interaction. Save the app version and any recent patch before reinstalling anything.

Do not run a long stress test to “prove” the graphics card is bad. A failing fan, high temperature, unstable power, or marginal setting can make repeated heavy testing less safe and less useful.

It affects the whole desktop or several apps

That widens the scope beyond one program. Record whether the mouse pointer, Task Manager, screenshots, sign-in screen, and a second display are affected. Then change one low-risk variable at a time.

Use the screenshot comparison carefully

Take a normal screenshot while the pattern is visible, then view that file on another known-good screen or phone. If the corruption appears inside the saved image, it entered the image before the display showed it, so the app, driver, rendered output, or graphics path remains in scope.

If the screenshot is clean but your photo of the monitor shows the problem, the cable, port, monitor, panel, or a display-only timing path becomes more plausible. A screenshot can narrow the path, but it cannot name the failed component by itself.

Compare one display path at a time

• One cable: power the computer down if the connector requires it, then compare one known-good compatible cable. Do not use a bent, hot, loose, or visibly damaged cable.
• One port: compare another supported output without forcing an adapter or connector. Record which port was used.
• One display: test a known-good compatible monitor or TV, or use the display’s built-in self-test when its manufacturer documents one.
• One dock or adapter: bypass it only when the computer and display support a direct connection. A connector that fits does not guarantee the same video capability.

Keep resolution and refresh rate within the display and cable’s documented support. Do not keep swapping hardware while the system is running unless the exact interface is designed for it.

Return only supported changes to stock

AMD’s current stability guidance recommends fact-finding first, factory-supported settings, and a record of each change. If the computer, graphics card, memory, or software has an overclock, undervolt, custom profile, riser cable, or third-party performance tool, note it.

Restore a setting only through the device maker’s documented process and only if you understand what will change. Do not flash firmware, change voltage, open a power supply, or remove internal hardware as a generic artifact test.

If a graphics-driver change came immediately before the symptom, save the current version and the prior known-good version. Use the computer or graphics maker’s support path. Avoid third-party driver sites and “driver cleaner” tools that make broad changes without a clear recovery plan.

Stop testing when heat or power becomes part of the story

Also stop if the screen is too corrupted to save important accessible files safely. Data preservation comes before another driver change or benchmark.

Avoid the repairs that create more damage

Do not bake a graphics card, use an oven or heat gun, attempt an improvised reflow or reball, add pressure to chips, bridge contacts, change voltages, or follow a board-level soldering video. A temporary visual change after heating is not a safe diagnosis, and uncontrolled heat can warp the board or damage nearby parts.

Do not buy a graphics card, monitor, power supply, or motherboard from the pattern alone. A useful assessment should explain which comparison ruled each part in or out.

Build a short evidence card for repair

• Computer, graphics card, dock, cable, and display make and model.
• The visible pattern and whether it appears before Windows, at sign-in, on the desktop, or in one app.
• Whether it appears in a screenshot viewed on another device.
• Results with one known-good supported cable, port, or display.
• Driver provider and version, Windows version, and recent app, update, repair, or hardware changes.
• Whether the system uses stock settings, custom profiles, a riser, adapters, or a dock.
• Any heat, fan, crash, shutdown, smell, connector, or power clues.

This record helps a technician reproduce the symptom without starting from a dramatic photo and guessing at the most expensive part.

When a physical inspection makes sense

Inspection is reasonable when corruption persists before Windows, returns across supported displays and cables, continues at documented stock settings, comes with heat or fan concerns, follows impact or liquid, or prevents reliable use after a careful driver and app review.

Budget Repair’s computer and laptop repair page explains the broader service path. You can also browse the repair guide library for related display and startup checks.

If you want an assessment, contact Budget Repair with the evidence card. We can inspect the display and graphics path, but we will not claim the GPU, video memory, monitor, cable, power supply, or motherboard failed from the image alone.