Artifacts on screen are the most expensive form of abstract art. At first it’s just a scattering of colored squares on top of the desktop, but after a couple of minutes it becomes clear that you’re not looking at a driver glitch, but at the cost estimate for the next month.
Just such a screenshot with a picture crumbling into pixelated mush was published by one of the PC owners with a caption in the spirit of “not now, not in this economy.”
The moment is really unfortunate: prices for video cards are far from normal, and after them, RAM has flown into space, where a set of two modules in some places costs as much as an entire console. In the comments, the diagnosis was made quickly and without painkillers – the video card was ready to retire.
The mechanics of dying are quite prosaic. The graphics chip and memory chips are held on the board not by wires, but by an array of tiny solder balls (BGA assembly), and any mechanical deformation of the PCB works to break these contacts. Most often, the memory chips located closest to the PCI-Express slot are the first to fall off, sometimes together with the contact pads, which are torn out of the board.
Hence the main advice, which in the discussion sounded harsher than any instructions from the manual:
Never allow your graphics card to flex. Even a little bit.
The problem became aggravated exactly when a two-slot card was no longer considered something impressive: modern three-four-slot bricks with massive heatsinks hang on the slot and one mounting screw, slowly folding under their own weight.
Any support is needed, even semi-legal methods – someone admitted that for many years now they have been using a crumpled plastic bottle as a support, calling it a “temporary permanent solution.”
The second killer is slower and does not depend on deflection at all. Each gaming session is a cycle of heating and cooling, the solder expands and contracts, and after a few years, metal fatigue takes its toll; This is exactly how at one time the Xbox 360 was massively mowed down with the famous “red ring of death”.
In laptops, where there is nothing to bend, degradation is added within the crystals themselves and the delayed consequences of static discharges, which can shoot out years later.
Gamers began to resell the RTX 5090, which they managed to buy before the insane rise in prices for top video cards
Wheel of Samsara for PC gamers – new technologies slow down old powerful video cards
The good news is that artifacts don’t always mean a funeral, and there’s a short checklist to go through before purchasing a new card:
-
Reinstall the card into the slot, cleaning the contacts and connector with isopropyl alcohol to remove dust and oxides
-
Replace the cable and try another monitor port, broken HDMI gives a similar picture
-
Completely remove the drivers via DDU and install a different version, preferably not the one that Windows slipped in
-
Try undervolting or lowering frequencies; sometimes semi-working memory stabilizes at a lower mode
Software is a different story. Windows 11 has long been known for its habit of overwriting manually installed AMD drivers with its version, and to this are added conflicts with proprietary utilities like Gigabyte Control Center and crashes in game video recording services.
Before you mourn the hardware, you should make sure that it is not the operating system that is to blame for deciding to take care of the user.
But let the advice to “warm up the card in the oven” remain where it belongs, in the folklore of the 2000s: warming up the BGA is done with a hairdryer, a soldering station and the hands of a person who knows what he’s doing, otherwise the repair will last exactly until the next session.
