The worst GPUs of all time, ranked

Published Jul 17, 2026, 9:30 AM EDT Monica J. White is a journalist with over a decade of experience in covering technology. She built her first PC nearly 20 years ago, and she has since built and tested dozens of PCs. PC hardware is her main beat, and graphics cards and the GPU market at large are her main area of interest, but she has written thousands of articles covering everything related to PCs, laptops, handhelds, and peripherals. From GPUs and CPUs to headsets and software, Monica's always willing to geek out over all things related to computing. Outside of her work with How-To Geek, Monica contributes to TechRadar, PC Gamer, Tom's Guide, Laptop Mag, SlashGear, Whop, and Digital Trends, among others. Her ultimate goal is to make PC gaming and computing approachable and fun to any audience. Monica spends a lot of time elbow-deep in her PC case, as she's always making upgrades, testing something, or plotting out her next build. She's the go-to tech support person in her immediate circle, so she's never out of things to do. Whenever she has spare time, you'll find her gaming until the early hours and hanging out with her dog. It's not even a debate: GPUs are the most exciting component in any PC. That's been the case for years and years. That only makes it all the more unfortunate when the graphics card is just plain bad ... and we've seen some really bad GPUs over the last three decades. This is the list of the worst GPUs of all time, and it won't be anything like what you'd expect. 7 Nvidia GeForce RTX 5090 The fastest GPU that's the hardest to recommend Right now, the RTX 5090 is the absolute best consumer graphics card in the world. There's no better. Nothing even comes near it. So why is it on this list? It's simple: it's impossible to recommend to just about anyone right now. The RTX 5090 started out at an MSRP of $1,999. That was hard to swallow, but these days, you'll find it selling for closer to $4,000-$5,000. This kind of price point pretty much invalidates any performance records the RTX 5090 managed to beat, but it's had other problems, too. Availability fluctuates, and for a time, it was pretty much just point-blank unavailable in stores. Like the RTX 4090, the 5090 has also seen cases of bent, melted connectors. But honestly, nearly all of that could be somewhat excused (after all, it's not every connector that melts) if the RTX 5090 wasn't so ridiculously overpriced. Thank the RAM-pocalypse. Asus GeForce RTX 5070 Ti Prime Memory Clock Speed 2,482MHz Graphics RAM Size 16GB The RTX costs over $4,000-$5,000, so I'd leave it alone. If you need higher-end performance, I wouldn't go above the RTX 5070 Ti these days. Nvidia GeForce RTX 5060 Ti The RTX 5060 Ti is a much cheaper alternative that still gets you plenty of VRAM, although spread across a narrow memory bus. 6 Nvidia GeForce GTX 480 The GPU you could use to fry eggs In 2010, Nvidia had a problem. AMD had already launched the Radeon HD 5870, the first DirectX 11 GPU, and Nvidia spent a whole six months preparing its response. That response was Fermi, and the flagship GTX 480 was meant to put Nvidia firmly back on top. It was fast, I'll give it that, which is why I'm fully expecting some pushback in the comments. It could beat the HD 5870 in many games, and its architecture was especially strong at tessellation and compute workloads. The problem was everything Nvidia had to do to achieve that performance. The GTX 480 had a 250-watt maximum board power rating, which is pretty normal these days but not so much back then. It consumed loads more power than its AMD rival. And all that power turned into a ridiculous amount of heat. In one contemporary test, the GPU reached 94 degrees Celsius under load, which translates to 201 degrees Fahrenheit. The exposed portion of the heatsink hit 67 degrees Celsius. As a result, the blower fan had to work overtime to keep it there, producing intense amounts of noise. 5 ATI Radeon HD 2900 XT The enormous flagship that couldn't catch Nvidia The Radeon HD 2900 XT arrived at a terrible time for AMD. Nvidia's GeForce 8800 series had already been on the market for a while, giving Nvidia a huge head start in the brand-new DX10 era. AMD needed a convincing response, and on paper, the HD 2900 XT sounded perfect. It was ambitious, imposing, and designed to make a statement. I miss seeing that in today's GPUs. Unfortunately, the massive specs didn't translate into the flagship performance buyers expected. It generally didn't beat the GeForce 8800 GTX that it was designed to defeat, and in its attempts, it produced lots of heat. Performance also fell apart once anti-aliasing entered the picture. AMD essentially spent months preparing an aggressive behemoth of a flagship only to fall short of Nvidia. It was a disappointment. 4 AMD Radeon RX 6500 XT Every corner AMD cut made the next one worse Launched amidst the awful GPU shortage of 2021-2022, the RX 6500 XT felt like an affordable savior met by a starved audience. But, even at $199, this GPU was pretty darn bad. The RX 6500 XT initially came with just 4GB of VRAM across a tiny 64-bit memory interface, using four PCIe lanes, and lacking H.264 and H.265 encoding. It was actually based on a laptop chip (Navi 24), and its limitations in desktops were hard to justify. The RX 6500 XT was outperformed by older cards like the RX 570, RX 590, and even the GTX 1650 Super, especially in PCIe 3.0 systems. AMD meant well, but the outcome was downright sad. 3 Nvidia GeForce GT 1030 DDR4 Two completely different GPUs under the same name The original GeForce GT 1030 was far from exciting, but it had some sort of a purpose as a cheap, low-power GPU that could fit a compact PC. With GDDR5 memory, it was even fast enough to make a little bit of sense for low-end gaming. Then Nvidia and its partners released another GT 1030 with DDR4 memory. The cards shared the same basic name and still came with 2GB VRAM, so many buyers naturally assumed that they were the same thing. But there was a huge gap between the GDDR5 and the DDR4 model: the GDDR5 offered around 48GB/s in bandwidth, while the DDR4 version dropped to roughly 16GB/s. Some also came with lower clock speeds. To the surprise of no one, in testing, the DDR4 version offered less than half the performance of the GDDR5 model. This was a good example of why a brand should never release two different products under the same name. 2 S3 ViRGE The infamous decelerator In the mid-1990s, dedicated 3D graphics were still a new idea, and most PCs relied on ordinary 2D graphics hardware, while games rendered their 3D visuals on the CPU. S3 was already one of the biggest names in PC graphics, so its ViRGE chip appeared to offer the perfect upgrade: strong 2D performance and hardware-accelerated 3D in one affordable package. The 2D part was genuinely great, but the 3D accelerator was actually referred to as a "3D decelerator." It could accelerate simple 3D scenes, but enabling more advanced features destroyed its performance. In some situations, rendering the game entirely on the CPU could actually be faster. This GPU wasn't wholly bad, but when the whole selling point was 3D acceleration and it failed at that, everything else sort of fell apart. 1 Nvidia GeForce FX 5800 Ultra The Dustbuster earned every part of its reputation Credit: Wikimedia Commons Can you believe that Nvidia once made a video mocking its very own GPU? That GPU was the FX 5800 Ultra, often nicknamed Dustbuster. The GPU arrived late, relying on extremely fast and hot-running GDDR2 memory across a 128-bit interface. It required an enormous dual-slot cooling system called FX Flow. The cooler pulled air through the card and exhausted it out of the case, and it was genuinely so loud that Nvidia itself compared it to a Harley Davidson and a hairdryer. All that awful noise might have been forgivable if the FX 5800 Ultra topped the charts, but it didn't. It narrowed the gap with ATI (now AMD), yet still struggled to consistently beat the Radeon 9700 Pro. It's remembered for its noise levels, not its performance. They may have been bad, but they win in one aspect As we move from one generation of GPUs to the next, we forget most of the cards. Incremental upgrades grow less impressive with each gen, and we mostly rely on DLSS and other AI-based features these days, anyway. But these GPUs won't easily be forgotten, even if they're remembered for how bad they were. Intel Arc B580 Memory Clock Speed 2,760MHz Graphics RAM Size 12GB Believe it or not, the modest Intel Arc B580 is actually a solid buy if you play games at 1080p.