Before 2020, we had little reason to create the list of top graphics cards for gaming. That list would’ve almost entirely consisted of Nvidia graphics cards, and would only be separated by price, availability, brand version, and value-for-money. Well, we’re glad to report that from 2020 onward, AMD rose to the challenge in the high-end market. Even better, they began dominating some portions, especially the mid-range and the budget end of it. For that reason, you’ll see a variety of models at vastly different price ranges in the best gaming graphics card list below. Let’s begin.
Note: Due to limited availability and unpredictable pricing, we’ll compare stock versions of every GPU model.
Nvidia GeForce RTX 3090 is the best gaming GPU money can buy. It has 10,496 GPU cores, a memory clock of 19.5 Gbps, and 24 GB of GDDR6X VRAM, and its core clock can fluctuate between 1.4 GHz (base) to 1.7 GHz (boost). As for power, it requires 2x 8-pin PCI-e cables and comes with a 350W TDP. Even better, its HDMI 2.1 can finally handle 8K gaming at 60 FPS in select games, although you’ll most gamers look to play in 4K at between 60 FPS and 120 FPS, depending on whether DLSS and Ray Tracking are enabled. Simultaneously, the GPU dominates creative workload tasks and is the only GeForce Ampere GPU with support for NVLink.
Navi 21 or AMD Radeon 6900 XT is the best graphics card for gaming this manufacturer has to offer. It comes with 5,120 RDNA cores, 128 MB of AMD Infinity Cache, a base clock of 1,825 MHz, a boost clock of 2,250 MHz. Additionally, its compute performance/memory speed is 23.04 TFLOPs and 16 Gbps respectively. And, while it’s true that it “only” has 16 GB GDDR6 VRAM, it comes at 1/3 of the MSRP of RTX 3090 while matching performance in (some) games in 4K resolution. It also stays below 300W TDP, has 80 Compute Units, and can handle 8K 60 FPS via HDMI 2.1, at least in theory.
AMD Radeon RX 6800 XT is our runner-up for the best GPU for gaming from AMD. The GPU’s performance is between 10% and 20% lower than the one above, but so is its MSRP – by 25%. The memory frequency stays high, between 2015 MHz (average in-game) and 2250 MHz (boost clock), and so does compute performance (20.74 TFLOPs), and AMD Infinity Cache (128 MB). Other things remain unchanged too – 300 WW TDP, 16 Gbps memory speed, and 16 GB of GDDR6 VRAM. The primary reduction in performance comes from 10% fewer Compute Units, 72, and slightly fewer RDNA cores, 4608.
Nvidia GeForce RTX 3080 is a top gaming graphics card for those with deep pockets but a reasonable budget. While it’s approximately 30% slower than 3090 in creativity workloads, its gaming performance only shows a reduction of 10%-15%. In contrast, its price is more than two times lower. You’re still getting faster VRAM than AMD’s contender, at 16 GB GDDR6X speeds. It also crushes ray-tracing and DLSS performance, thanks to a combination of 8,7004 CUDA cores, and updated Tensor Cores. And, although its memory clocks look low at 1,440 MHz base and 1,710 MHz boost clocks, its compute performance is 29,76 TFLOPs.
If you think the lack of the “XT” part is the main difference, you’re wrong. In reality, AMD Radeon RX 6800 is the gaming GPU with the best value-for-money ratio e.g., FPS per dollar. Its MSRP dropped by a further 7% compared to RX 6800 XT, and so did its performance in theory. But in reality, you’ll still get plenty – 20,74 TFLOPs performance, 60 Compute Units, and 3,840 RDNA cores. The GPU also uses RDNA 2 architecture and has a base clock of 1,700 MHz, a boost clock of 2,105 MHz, and 16GB GDDR6 VRAM. On the other hand, a drop in TDP to 250 W means there’s room for overlocking.
Both Nvidia GeForce RTX 3060 and RTX 3060 Ti are valid competitors alongside RX 6800, and each has a right to be called the “best bang for the buck graphics card”. With this one, you’re getting RTX 2080 performance at a much lower price, perfect to simulate a console gaming experience, e.g., 4K@30FPS. Only this time, you can enable High graphics, Ray Tracing, and DLSS. Finally, we ask you to remember these numbers and compare them in a jiffy. Those are – number of CUDA cores, 4,864, base clock, 1,410 MHz, boost clock, 1,665 MHz, and compute performance, 12,74 TFLOPs.
We know the list looks repetitive. The truth is, the performances and the value-for-money values are wildly different. As such, each GPU has a portion of the market it dominates. For example, Nvidia GeForce RTX 3070 is a leading 1440p gaming graphics card that’s also gunning for the 4K@30 FPS gaming fans. It essentially provides RTX 2080 Ti performance at half the price, while still being 20% more expensive than RTX 3060. It’s also smaller while being more powerful, at 5,888 CUDA cores, 8GB GDDR6 VRAM, a base clock of 1,500 MHz, and a boost clock of 1,725 MHz.
Now’s time for AMD’s contender for the throne of a 1440p gaming leader. The MSRP is nearly identical to the one above. But, with Ray Tracing and DLSS enabled, RTX 3070 destroys RX 6700 XT by close to 170% in 1080p and 200% in 1440p. Obviously, that’s not its specialty, and not what it’s intended for. It provides a solid number of RDNA cores, 2,560, and 50% more VRAM than RTX 3070 – 12GB GDDR6 to be exact. It’s also lighter and smaller while being equally energy-efficient. Finally, it comes with a 2,30 GHz core clock and a 2,58 GHz boost clock.
Nvidia GeForce RTX 2070 Super is the most affordable 1440p gaming graphics card from our analysis. It’s also widely available, unlike the others above. It has 2560 CUDA cores, 14 Gbps memory speed, 8 GB GDDR6 VRAM, and a clock frequency of 1605 MHz (base) and 1,770 MHz (boost). At the same time, it uses less power than any Nvidia GPUs above and produces less heat at a TDP of 215 W. The card even supports NVLink, so you can pair two of them in SLI even though games that truly benefit from SLI are scarce.
Don’t fret, budget gamers, we haven’t forgotten you. Nvidia GeForce GTX 1660 Super is the best cheap GPU for gaming in 1080p. It’s a perfect choice if you’re strapped for cash, especially if you’re stuck with a 60 Hz -144 Hz gaming monitor. And, even though it doesn’t feature Ray Tracing nor DLSS technology, it does have 1,408 CUDA cores, 6GB GDDR6 VRAM, 14 Gbps memory speed, and base/boost clocks of 1,530/1,785 MHz respectively. It’s also energy-efficient and doesn’t get too warm thanks to a 125 W TDP. The GPU doesn’t even need a powerful PSU, since it uses a single 8-pin PCIe connector.