Random coloured dots, flickering triangles, lines across the screen, or textures that look like they have melted. These are artifacts, and they mean the graphics card is producing incorrect output. Unlike most computer symptoms, they point at one component fairly reliably.
The short answer
Check the temperature under load first, because overheating is the most common and most fixable cause. Then clean the dust out and confirm the card’s power connectors are properly seated with separate cables. Artifacts that appear only in games point at heat or power; artifacts on the desktop and in the BIOS point at the card itself failing.
Where the artifacts appear tells you a lot
- Only in games or heavy 3D work. The card under load: heat, power delivery, or an unstable overclock. Usually fixable.
- On the desktop too. More serious. The card is failing even at low load.
- In the BIOS, before Windows loads. Definitively hardware. No driver or software can be responsible at that point.
- Only in one game. Likely that game or its shaders, not the card. Verify the game files and update the driver.
- After a driver update. Roll it back and reinstall cleanly before assuming hardware.
- Accompanied by crashes to desktop. Driver instability, often heat-related.
Monitor the temperature under load
Run something demanding for twenty minutes with monitoring software open. Graphics cards run hot by design, and up to about 85 degrees Celsius is normal. Sustained above 90, or memory temperatures above 100 on cards that report them, is where artifacts start.
Clean the card thoroughly
Power down, unplug, and clear dust from the card’s heatsink fins and fans. Hold the fans still while using compressed air. A card that has been in service for years is often visibly clogged, and this alone resolves a lot of thermal artifacts.
Check the power connectors
Make sure each supplementary power connector is fully seated and clicked home. Critically, use separate cables from the power supply rather than one cable with two daisy-chained connectors, which is a common cause of instability under load.
Remove any overclock
Set the card back to stock clocks and memory speeds, including any factory overclock profile in the manufacturer’s software. An overclock stable for a year can become unstable as the card ages.
Reinstall the driver cleanly
Use the manufacturer’s removal tool to strip the existing driver completely, restart, then install the current version fresh. Installing over the top leaves fragments that cause exactly this.
Test in the BIOS
Boot into the BIOS and leave it there for a few minutes. Artifacts visible in the BIOS rule out drivers and software entirely, which is a definitive result worth having.
Reseat the card
Power down, remove the card from its slot, check the contacts are clean, and refit it firmly. A marginal seating causes intermittent faults, and this costs nothing.
Test the card in another machine if you can
The definitive test. If artifacts follow the card, the card is at fault. If they stay with the machine, look at the power supply or the motherboard slot.
Underclock the card as a stopgap if you need the machine working. Reducing the core and memory clocks below stock, using the manufacturer’s software, often eliminates artifacts on a card with degrading memory. It costs performance and it can keep a machine usable while you arrange a replacement.
Memory degradation
The most common failure mode on an ageing graphics card is the memory rather than the processor. Video memory runs hot, particularly on cards where the memory chips are on the back of the board with limited cooling, and it degrades over years of heavy use.
The signature is artifacts that appear in memory-intensive situations first: high resolutions, high texture settings, or games with large worlds. Reducing texture quality and resolution often makes them disappear, which is itself diagnostic. Underclocking the memory specifically, rather than the core, frequently helps more.
Cards used heavily for cryptocurrency mining often have degraded memory and worn fans, and they turn up on the second-hand market looking fine. If artifacts appeared on a card you bought used, that history is a likely explanation.
When it is not the graphics card
- A failing power supply. Cannot hold voltage under load, and the card glitches. If the machine also restarts under load, suspect this strongly.
- A bad video cable. Usually produces regular patterns, colour tints or dropouts rather than the random shapes of true artifacts. Swap it as a cheap test.
- The monitor. Test with a different display.
- System memory. Can produce visual corruption too. A memory test is worth running.
- An overheating motherboard chipset, rare but possible.
What are graphics artifacts?
Incorrect output from the graphics card: coloured dots, flickering shapes, lines, or corrupted textures. They mean the card is producing wrong data rather than the display showing it wrongly.
Does seeing artifacts mean my card is dying?
Not necessarily. Heat, dust, power delivery and an unstable overclock all cause them and all are fixable. Artifacts visible in the BIOS are the ones that point firmly at hardware failure.
What temperature is too hot for a graphics card?
Up to about 85 degrees Celsius under load is normal. Sustained above 90, or memory temperatures above 100 where reported, is where problems begin.
Can a power supply cause artifacts?
Yes. A supply that cannot hold voltage under load makes the card glitch. Suspect it particularly if the machine also restarts unexpectedly under load.
Can I keep using a card that artifacts?
Underclocking it often removes the artifacts at a cost in performance, which keeps the machine usable while you arrange a replacement. It is a stopgap rather than a fix.
The bottom line
Check temperatures under load and clean the dust out, because heat is the most common cause and the easiest to address. Confirm the power connectors are seated with separate cables, and remove any overclock. Artifacts visible in the BIOS rule out software entirely, and underclocking is a legitimate stopgap for a card with degrading memory.
