Nvidia GeForce RTX 4090: The $1,600 power connector that melted
When Nvidia launched the RTX 4090 in late 2022, it was the fastest consumer GPU ever made. But within weeks, reports started flooding forums: the new 12VHPWR power connector was literally melting. The problem? A poorly designed adapter that could overheat if the cable wasn’t seated perfectly — or if it was bent too sharply inside a case.
Nvidia eventually blamed user error, but the damage was done. For a flagship card that cost $1,600 at launch, this wasn’t just a defect. It was a worst GPU moment in an otherwise stellar generation. The RTX 4090 remains a beast for gaming and AI workloads, but its launch will forever be tied to fire hazards and RMA horror stories.
The connector fiasco in numbers
Over 100 confirmed cases of melted connectors were documented by Gamers Nexus alone. Nvidia revised the adapter months later, but the trust was gone. If you’re building a high-end PC today, you’re still gambling on that 12VHPWR plug — and that’s a problem Nvidia has yet to fully solve.
Intel Arc A770: The comeback that never was
Intel entered the discrete GPU market in 2022 with the Arc A770, promising to challenge Nvidia and AMD. The hardware was ambitious: 16GB of VRAM, ray tracing support, and XeSS upscaling. But the drivers were a disaster. Games that ran fine on any Nvidia or AMD card would crash, stutter, or refuse to launch on Arc.
Intel fixed many of the driver issues over the following year, but for early adopters, the Arc A770 was a frustrating, unfinished product. By the time it worked well, buyers had already moved on. It’s a textbook case of why worst GPUs aren’t always slow — sometimes they’re just broken.
AMD Radeon RX 7900 XTX: The vapor chamber that failed
AMD‘s Radeon RX 7900 XTX launched in December 2022 as a flagship competitor to Nvidia’s RTX 4080. But within weeks, reviewers and users noticed something alarming: the card would overheat and throttle under load, with GPU hotspot temperatures exceeding 110°C. The culprit was a defective vapor chamber in some reference models.
AMD acknowledged the issue and offered replacements, but the damage to the RX 7900 XTX’s reputation was real. For a $1,000 card, this was unacceptable. The worst graphics cards in history share a common thread: they fail at the basics. The 7900 XTX failed at staying cool.
Nvidia GeForce RTX 4060 Ti: The 8GB embarrassment
In May 2023, Nvidia released the RTX 4060 Ti with 8GB of VRAM — at a time when games like Hogwarts Legacy and The Last of Us Part I were already exceeding 8GB at 1080p. Gamers were furious. Why would Nvidia ship a $400 card with such limited memory?
The answer was obvious: market segmentation. Nvidia wanted to push buyers toward the more expensive 16GB version (which cost $500 and barely sold). But the 8GB RTX 4060 Ti became a punchline. It’s one of the worst GPUs of all time not because it’s slow, but because it’s insulting.
Nvidia GeForce RTX 4080 12GB: The card that got cancelled
Originally announced as the RTX 4080 12GB alongside a 16GB version, this card caused immediate backlash. The two “4080” models had completely different specs: different memory buses, different core counts, and different performance levels. Gamers called it deceptive, and Nvidia quickly backtracked.
In October 2022, Nvidia “unlaunched” the 12GB model, rebranding it as the RTX 4070 Ti months later. The whole episode was a PR disaster. It’s a rare case of a worst graphics card that never even hit shelves — but the confusion and anger it caused still haunt Nvidia’s reputation.
AMD Radeon RX 6500 XT: The 4GB card that couldn’t even play modern games
Released in January 2022, the RX 6500 XT was supposed to be an entry-level card for budget gamers. But AMD cut so many corners that it was nearly unusable. It had only 4GB of VRAM — at a time when even $200 cards had 6GB or 8GB. Worse, it lacked hardware encoding for video streaming and had no 4K output support.
In benchmarks, the RX 6500 XT was slower than the RX 570 from 2017. It couldn’t run Call of Duty: Warzone at 1080p without stuttering. It’s widely considered one of the worst GPUs of the modern era, and a low point for AMD’s Radeon division.
Nvidia GeForce GTX 970: The 3.5GB lie
The GTX 970 was a popular mid-range card when it launched in 2014. But users soon discovered a dirty secret: Nvidia advertised it as having 4GB of VRAM, but the last 0.5GB was accessed at a much slower speed. In practice, the card performed like it had 3.5GB of fast memory and 0.5GB of slow memory.
The class-action lawsuit that followed forced Nvidia to pay out $30 million in refunds. The GTX 970 is a classic example of a worst GPU that sold millions — but left a bitter taste for everyone who bought one.
Intel Arc A750: The driver nightmare continues
Intel’s second attempt at a discrete GPU, the Arc A750, launched alongside the A770 in 2022. It was cheaper, but it inherited the same driver problems. Games that relied on older DirectX 9 or 11 APIs ran poorly or not at all. Intel promised fixes, but they came slowly.
By the time the drivers were stable — around mid-2023 — the A750 was already obsolete in terms of performance. For budget-conscious buyers, it was a gamble that rarely paid off. The A750 is another entry on the list of worst graphics cards that could have been great if Intel had just waited until the software was ready.
Nvidia GeForce RTX 3080 12GB: The paper launch that angered everyone
In January 2022, Nvidia released an updated RTX 3080 with 12GB of VRAM and a slightly faster core. But stock was virtually nonexistent. Scalpers bought up every unit, and the card was sold at double its $699 MSRP for months. Gamers who couldn’t find a standard RTX 3080 were left frustrated.
The 12GB version was a paper launch — a product that existed mostly on paper and in press releases. It’s a reminder that the worst GPUs of all time aren’t always the slowest; sometimes they’re the ones you can’t even buy.
Nvidia GeForce RTX 4060: The 60-class card that barely improved
When Nvidia released the RTX 4060 in June 2023, expectations were low. The RTX 3060 had been a solid 1080p card, and the 4060 needed to beat it. It did — by about 15%. But that was only with DLSS 3 Frame Generation enabled. Without it, the 4060 was barely faster than the 3060, and sometimes slower in ray tracing.
At $299, it wasn’t a bad deal. But for a generation-to-generation upgrade, it was disappointing. The RTX 4060 is a worst GPU in the sense that it represents everything wrong with modern GPU launches: tiny performance gains, reliance on upscaling, and a lack of meaningful innovation.
The common thread: Why these GPUs failed
Whether it’s melting connectors, broken drivers, or deceptive marketing, the worst graphics cards in history share one thing: they broke the trust between manufacturers and users. A GPU is the most expensive component in a gaming PC. When it fails — literally or metaphorically — it leaves a lasting impression.
If you’re shopping for a new GPU today, learn from these mistakes. Read reviews from multiple sources. Check for known issues like connector problems or driver bugs. And never, ever buy a card based on promises of future fixes. The best GPU is one that works out of the box.
For more on avoiding bad hardware, check out our guide on how to choose a graphics card and our ranking of the best GPUs for gaming in 2025.
