GPU Artifacts: What They Look Like and What Causes Them

GPU artifacts are visual anomalies that appear on screen when the graphics card outputs corrupted or incorrect data. They range from a fixable driver problem you can solve in ten minutes to a sign of permanent VRAM damage that no software can reverse. The difference between those two outcomes is not luck — it is understanding what type of artifact you are dealing with and what is causing it. This guide covers every category: what each looks like, what causes it, the temperature thresholds involved, and how to diagnose which situation you are actually in. If you are troubleshooting a broader hardware problem, our complete PC troubleshooting and hardware diagnostics guide is the place to start before going deep on any single component.

What GPU Artifacts Actually Look Like

GPU artifacts are visual anomalies caused by incorrect or corrupted data from the graphics card, appearing as pixel blocks, screen flickering, texture corruption, or geometric distortion.

Before diagnosing the cause, you need to identify what you are seeing. The visual form of an artifact gives you strong clues about the origin:

Pixel blocks and colored squares — Small clusters of wrongly colored pixels, often appearing as bright green, pink, or white squares scattered across the display. These are the most common artifact type and almost always indicate either VRAM instability or a driver issue.

Texture corruption — In-game textures that stretch, repeat incorrectly, appear as solid color patches, or display with entirely wrong colors. The geometry of the scene looks fine but the surfaces look broken. This is typically a VRAM read/write error.

Geometric spikes and triangle tears — Polygons or triangles that stretch to extreme lengths across the screen, sometimes filling the entire display. These geometry artifacts usually indicate the GPU core is struggling, not VRAM — often seen with severe overclocking or dying shader hardware.

Screen flickering — The entire display or sections of it flash on and off rapidly. Can be caused by driver issues, a loose cable, a failing display output on the GPU, or power delivery problems.

Color banding and rainbow patterns — Gradients that should be smooth showing visible color steps, or strange repeating color patterns. More common with driver corruption or display cable issues than with VRAM damage.

Persistent pixel corruption at desktop — Artifacts that appear even on the Windows desktop, in the BIOS, or during POST. This is the most serious category. If artifacts appear before any driver has loaded, the hardware itself is damaged.

The context in which artifacts appear — under gaming load only, in specific applications, at desktop, or during boot — is the single most diagnostic piece of information you have.

The 5 Root Causes of GPU Artifacts

After diagnosing hundreds of artifact complaints across the machines we monitor, the split is roughly: 60% thermal, 25% driver, 15% hardware damage. Each category has a distinct fingerprint.

1. GPU Overheating or VRAM Overheating

Symptoms: Artifacts appear under sustained load (gaming, rendering, stress testing) and disappear when the GPU cools down. Desktop is clean.

Temperature thresholds: GPU core hotspot temperatures above 95°C consistently produce instability on most Ada Lovelace (RTX 40-series) and RDNA 3 (RX 7000-series) cards. GDDR6X VRAM junction temperature — the memory chip temperature, not the core — is rated by Micron to a maximum of approximately 105°C, with thermal throttling and memory errors appearing at sustained junction temperatures above 95-100°C. In practice, GDDR6X cards like the RTX 4090 and RTX 4080 run VRAM junction temperatures of 80-88°C under normal load. Cards with degraded thermal pads — the material between the VRAM chips and the heatsink — can push junction temps to 100°C or above, at which point artifacts appear predictably.

Recoverable: Yes, if addressed before physical damage sets in. Replacing thermal pads on RTX 4090 cards, for example, typically drops VRAM junction temperature by 15-25°C.

Key diagnostic: Monitor VRAM junction temperature specifically, not just GPU core temperature. A card showing 75°C on the core sensor can simultaneously show 102°C on the VRAM junction if the thermal pads have dried out. Most GPU monitoring tools (HWiNFO64, GPU-Z) expose this sensor.

2. Driver Corruption or Incompatibility

Symptoms: Artifacts appear in most or all applications, often beginning immediately after a driver update. Artifacts are visible during games and in the Windows desktop, but NOT during boot or BIOS — the operating system has loaded before they appear.

Recoverable: Yes. Use Display Driver Uninstaller (DDU) in Safe Mode to completely remove the existing driver, then install either the previous stable version or the latest release from NVIDIA or AMD directly.

Key diagnostic: Roll back the driver first. If artifacts disappear, driver corruption was the cause. This takes ten minutes and costs nothing. Always eliminate driver corruption before assuming hardware failure.

Note on partial corruption: Drivers that are partially corrupted can cause artifacts in specific applications or only under specific rendering conditions. A fresh driver install using DDU eliminates this variable entirely.

3. Physical VRAM Damage from Sustained Overheating

Symptoms: Artifacts present at the Windows desktop, in the BIOS, during POST, or even as the machine boots. Not load-dependent. Pixel corruption does not disappear when temperatures drop. Getting progressively worse over weeks or months.

Recoverable: No. Physical VRAM cell damage is permanent. Individual memory cells that have failed due to sustained heat exposure do not recover. The GPU will continue to function with errors, degrading further over time, until failure becomes complete.

Temperature history: This damage accumulates over time. A card that ran at VRAM junction temperatures of 100°C+ for dozens of gaming sessions over months is far more likely to develop permanent artifacts than one that ran hot once. This is precisely why the pattern matters more than any single reading — something you can only see with a monitoring history, not a one-time manual check.

Key diagnostic: If the artifacts are visible before the OS loads (during BIOS splash, during POST), no driver fix will help. Test with a live USB or second OS if you need to confirm this.

4. Unstable Overclocking

Symptoms: Artifacts appear after increasing GPU memory clock or core clock. Dropping clock speeds back to stock eliminates the artifacts. Load-dependent but distinguishable from thermal artifacts because temps may be completely normal.

Recoverable: Yes — immediately and completely by resetting to stock clocks. However, if overclocking was combined with sustained thermal stress, some physical VRAM damage may have already occurred, which may not recover even after returning to stock.

Threshold: Memory overclocking artifacts typically appear when the memory clock exceeds what the VRAM can reliably sustain at its current voltage and temperature. Factory-overclocked cards (OC editions from ASUS ROG, MSI Gaming X, etc.) are binned for those speeds, but custom overclocking beyond factory settings immediately enters unstable territory for many units.

Key diagnostic: Reset to stock clocks in MSI Afterburner, ASUS GPU Tweak, or AMD Software Adrenalin. If artifacts vanish within a minute of running at stock, overclocking was the cause.

5. Insufficient or Degraded Power Delivery

Symptoms: Artifacts appear during heavy GPU load, often simultaneously with system instability, BSOD, or sudden shutdowns. May correlate with PSU age (3+ years under heavy load). Core and VRAM temperatures are normal.

Recoverable: Yes — typically by replacing the PSU or ensuring adequate PCIe power cable connections. RTX 40-series cards using the 16-pin (12VHPWR) connector have shown issues when connector seating is imperfect, causing power delivery drop under load.

Key diagnostic: Check PSU rail voltages under load using HWiNFO64 (12V rail should stay above 11.4V under full GPU load). Measure PSU load versus rated wattage — a 650W PSU powering a system drawing 700W peak will produce artifacts, instability, and eventually failure.

Thermal Artifacts: The Most Common and Most Recoverable

Thermal artifacts deserve extended coverage because they are both the most common category and the one where the window for intervention is clearest. For a deeper look at the full spectrum of GPU overheating symptoms, our post on 5 signs your GPU is overheating and how to prevent it covers the early warning signs before artifacts appear.

The thermal artifact cycle works like this: GPU core or VRAM heats up under sustained load → memory cells or shader units become thermally unstable → incorrect data gets written or read → visual corruption appears on screen → GPU throttles or restarts to protect itself → temperatures drop → artifacts disappear. This cycle repeating itself is the definition of a thermal artifact.

Two temperatures matter here, and conflating them is a common diagnostic mistake:

GPU Hotspot (Junction) Temperature: The highest temperature measured anywhere on the GPU die. On Ada Lovelace (RTX 40-series) cards, hotspot temps of 85-90°C are normal under load. Sustained hotspot temperatures above 95°C indicate inadequate cooling and will eventually produce core instability. NVIDIA specifies a maximum GPU junction temperature of 89°C for standard operation on the RTX 4090 — the card throttles beyond this to protect itself. See our detailed guide on GPU hotspot vs. edge temperature for the difference between these two readings.

VRAM Junction Temperature: The temperature of the VRAM chips themselves, exposed through a separate sensor. On GDDR6X cards (RTX 4090, RTX 4080, RTX 4070 Ti), Micron rates the memory to a maximum junction temperature of approximately 105°C. However, artifacts and memory errors reliably appear at sustained VRAM junction temperatures above 95-100°C. The thermal pads that transfer heat from the VRAM chips to the heatsink are the critical component here — they degrade over years of thermal cycling and lose conductivity, which directly raises VRAM junction temperature without any change in GPU core temperature.

Practical test: Run a GPU stress test (FurMark, OCCT, or 3DMark TimeSpy stress) for 20-30 minutes while monitoring both GPU hotspot and VRAM junction temperature simultaneously in HWiNFO64. If artifacts appear and VRAM junction is above 95°C, thermal pad replacement is the intervention. If artifacts appear with both temperatures under 85°C, thermal is not the primary cause.

For cards with degraded thermal paste on the GPU die itself (common on cards over 3 years old), reapplying thermal paste can drop GPU hotspot temperature by 10-20°C, often eliminating core-driven thermal artifacts entirely without touching the VRAM pads.

How to Diagnose GPU Artifacts: Step by Step

This is the sequence to follow every time, in order. Skipping steps wastes time and leads to misdiagnosis.

1. Note when and where artifacts appear. Desktop only? During gaming? During boot? In one application or all applications? Write this down before doing anything else. It is the most important diagnostic data you have.

2. Check temperatures under load. Open HWiNFO64 and run a GPU stress test. Record GPU hotspot temperature AND VRAM junction temperature. If hotspot exceeds 95°C or VRAM junction exceeds 95°C sustained, thermal is your primary suspect.

3. Reset overclocking to stock. If the card has any OC profile applied — factory or custom — reset it to stock clocks in your GPU tuning software. Run the stress test again. If artifacts disappear, overclocking was the cause.

4. Clean install GPU drivers. Download DDU, boot into Safe Mode, uninstall all GPU drivers completely, reboot, install the latest driver from NVIDIA.com or AMD.com directly. Run the stress test again. If artifacts disappear, driver corruption was the cause.

5. Test PSU load and voltage. In HWiNFO64, monitor the 12V rail voltage under full GPU load. If it drops below 11.4V, PSU inadequacy is a factor. Check that PCIe power cables are fully seated — on 12VHPWR adapters, all the way until you hear or feel the latch click.

6. Test with a different display cable and port. A failing DisplayPort cable or a dying GPU display output can produce flickering and color artifacts that look identical to VRAM issues. Swap to a different cable and a different port on the GPU. If artifacts change or disappear, the cable or output is the cause.

7. Run GPU memory diagnostic. Tools like MemTestG80 or OCCT's VRAM test specifically stress VRAM and report memory errors. A healthy card produces zero errors. Any errors confirm VRAM instability, either from heat or physical damage.

8. Observe artifacts at boot/BIOS. If you reach this step and artifacts still appear, look for them during the BIOS splash screen or POST. If visible before the OS loads, hardware damage is confirmed.

For related diagnostic patterns — particularly when the GPU is involved in crashes rather than just visual corruption — our step-by-step guide to diagnosing PC crashes under load covers the full diagnostic tree including GPU, PSU, and RAM interactions.

When Artifacts Mean the GPU Is Dying

Not all artifacts signal permanent failure. But some patterns are unambiguous indicators that the hardware is beyond software fixes:

The clearest signs of hardware damage:

• Artifacts visible in the BIOS or during POST (before any driver loads)
• Artifacts present on the desktop at idle, when GPU temperature is 40°C or lower
• Artifacts that do not disappear when clocks are reduced to minimum
• MemTestG80 or OCCT VRAM test reporting hundreds or thousands of errors
• Artifacts that have gradually worsened over weeks, not appeared suddenly
• Multiple pixel corruption patterns appearing simultaneously (not just one type)

The intermittent failure pattern: One of the more frustrating presentations is a GPU that artifacts intermittently for weeks before settling into permanent failure. VRAM cells under thermal stress do not all fail at once. Early VRAM damage produces occasional corruption under specific load conditions. As more cells fail, the threshold for triggering artifacts lowers until they appear constantly. This gradual progression from "artifacts under heavy load" to "artifacts everywhere" typically takes between 2 weeks and 3 months and is a reliable indicator that hardware damage has begun.

What cannot be fixed: Individual VRAM cells that have physically failed from sustained heat exposure cannot be repaired by consumer-available means. Reballing VRAM chips is possible in professional soldering environments but costs more than most mid-range cards are worth. For most users, permanent VRAM damage means GPU replacement.

What looks like dying but is not: A GPU that artifacts only during gaming, only at high temperatures, and clears up completely at idle or after a driver reinstall is almost certainly recoverable. These thermal and driver-driven artifacts are not hardware damage — they are hardware stress. The distinction matters enormously for deciding whether to clean the card, replace thermal pads, or buy a new GPU.

Understanding GPU temperature stress before it causes VRAM damage requires seeing the temperature history over time. Artifacts that arrive suddenly often have a weeks-long thermal history behind them — visible in monitoring data as a gradual upward drift in VRAM junction temperature as thermal pads degraded. The machine knew this was coming. Continuous GPU temperature monitoring catches thermal artifacts before they cause VRAM damage.

Prevention: How Monitoring Catches This Early

The pattern we see across the machines in our monitoring fleet is consistent: GPU artifact complaints typically arrive after weeks of sustained VRAM junction temperatures above 95°C visible in the monitoring history. The user notices the artifacts on a Tuesday. The data shows the VRAM junction temperature crossed 95°C under load for the first time four months ago, and has been climbing gradually since.

Manual spot-checks miss this entirely. A one-time temperature reading tells you nothing about trend. A monitoring history tells you when the problem started, how fast it progressed, and whether the intervention (thermal pad replacement, driver update, OC reset) actually worked.

GGFix monitors GPU hotspot temperature, VRAM junction temperature, GPU edge temperature, fan speed, and GPU load continuously — tracking the pattern, not just the snapshot. When VRAM junction temperature begins trending above 90°C sustained, the system generates an alert before the user has seen their first artifact. That is the difference between a 200 DKK thermal pad replacement and a 5,000+ DKK GPU replacement.

For the full picture of thermal management across all PC components — not just GPU — the thermal throttling explained guide covers how temperature-induced performance reduction works across CPU, GPU, SSD, and VRM simultaneously, and why monitoring each independently matters.

Frequently Asked Questions

Q: Do GPU artifacts always mean the card is dying?

No. The majority of GPU artifacts are caused by driver corruption, overclocking instability, or thermal stress — all of which are recoverable without hardware replacement. Artifacts that appear only under load and disappear when the GPU cools down, or that started after a driver update and stopped after a rollback, are not hardware failure. Artifacts that appear at the desktop, at idle temperatures, or during BIOS/POST are the ones that indicate physical hardware damage. Eliminate driver and thermal causes first before concluding the GPU is dying.

Q: At what temperature do GPU artifacts start appearing?

For GPU core hotspot temperature, artifacts typically begin appearing at sustained temperatures above 95°C on modern NVIDIA Ada Lovelace and AMD RDNA 3 cards. For GDDR6X VRAM junction temperature specifically, Micron rates the memory to approximately 105°C maximum, but memory errors and visual artifacts reliably appear at sustained junction temperatures above 95-100°C. Both sensors need to be monitored — a card showing 75°C GPU core can simultaneously have 102°C VRAM junction if thermal pads have degraded, and the VRAM junction is where the artifacts originate in that scenario.

Q: Can GPU artifacts be fixed by updating drivers?

Yes — if driver corruption is the cause. The diagnostic test is straightforward: use Display Driver Uninstaller (DDU) in Safe Mode to completely remove the existing driver, then install a fresh copy directly from NVIDIA.com or AMD.com. If artifacts appeared after a recent driver update, roll back to the previous version first. Driver-caused artifacts typically affect the OS desktop as well as games, appear in multiple applications, and started immediately after a driver change. If a clean driver reinstall eliminates the artifacts, no further action is needed.

Q: What is the difference between GPU artifacts and screen tearing?

Screen tearing is a specific visual effect where the display shows two or more frames simultaneously, producing a horizontal split line across the image. It is caused by the GPU rendering frames faster than the monitor's refresh rate, with no synchronization (VSync, G-Sync, or FreeSync) in place. Screen tearing is not a hardware fault — it is a synchronization mismatch and is fully resolved by enabling VSync or an adaptive sync technology. GPU artifacts, by contrast, involve corrupted or incorrect data: wrong pixel colors, distorted textures, geometric spikes, or persistent pixel corruption. Tearing has clean horizontal splits; artifacts look genuinely broken.

Q: How do I tell if artifacts are from the GPU or the display cable?

The key test is to swap the display cable and port. Use a different cable, connect to a different port on the GPU (HDMI instead of DisplayPort, or a different DisplayPort output), and if possible test on a different monitor. Cable-caused artifacts typically manifest as flickering, color shift, or intermittent signal loss rather than pixel blocks or texture corruption. They are also usually consistent regardless of GPU load. If swapping the cable and port eliminates the artifact, the cable or GPU display output is the cause, not VRAM. A cable issue is a 50-200 DKK fix; VRAM damage is not.

Q: Can overclocking cause permanent GPU artifact damage?

Yes, if the overclocking was combined with insufficient cooling over an extended period. Overclocking increases both power consumption and heat output. If a card was run at high memory clocks while also running hot — VRAM junction temperatures above 95°C sustained over many gaming sessions — the thermal stress accelerates VRAM cell degradation. Resetting to stock clocks can eliminate the overclocking instability component of the artifacts, but any thermal damage already accumulated to the VRAM cells is permanent. Overclocking on an adequately cooled card that maintains VRAM junction temperatures below 90°C carries much lower physical damage risk than overclocking a card that was already running hot.