Whether you’ve been playing PC games for years or you’re building your first gaming PC, picking the right graphics card is probably one of the most consequential choices. It doesn’t help that now there’s three companies making graphics cards, which means there are more options out there than ever before.
But there are a lot of reasons to pick an Nvidia graphics card over one made by AMD or Intel. For one, whether you love it or hate it, Nvidia has plenty of software and features that are exclusive to its graphics cards, whereas things like FSR (FidelityFX Super Resolution) from AMD will work no matter what GPU is in your system. Features like DLSS, or Deep Learning Super Sampling, make an Nvidia graphics card an attractive choice.
Nvidia also has the most powerful graphics cards on the market right now, especially at the high-end. There simply is not another graphics card that can match the RTX 5090 in pure performance. But even when it comes to more affordable graphics cards like the RTX 4060 Ti, the AI upscaling offered by DLSS can really help stretch the performance, and can even push into higher resolutions than you would otherwise be able to.
It's still important to figure out what resolution you want to play games at, because that’s going to change what type of graphics card you want to go for. If you want to play all the newest PC games at 4K with all the settings maxed, you’re going to need a much more powerful card than someone that just wants to play World of Warcraft at 1080p.
TL;DR: These Are Best GPUs
Graphics Cards Basics
While graphics cards are extremely complicated devices, shopping for one doesn’t need to be that much of a hassle. As long as you keep an eye out for some simple specs, you should have a pretty good idea of what you’re getting into.
The most obvious thing to look for is whether or not your graphics card is actually part of the current generation. You don’t want to miss out on performance or features, especially if you’re spending hundreds of dollars on just one part of your computer. With Nvidia graphics cards, this is actually extremely easy, just look at the two numbers following the ‘RTX’ or ‘GTX’ in the graphics card’s name. Nvidia just launched its newest generation of graphics cards, which are labeled with '50.' If it’s ‘40,’ that’s the previous generation from 2022; ‘30’ and ‘20’ are the two generations before that.
Don’t get me wrong, getting a 30-series or 20-series card will still provide an excellent gaming experience, with the RTX 3080 still able to power most games at 4K without breaking a sweat.
But what about the second two numbers? Well, take a look at the RTX 5090, the top-end graphics card on the market. Then, take a look at the RTX 4050, which is only a laptop GPU, but is way weaker. The scale typically goes from 50-90, with the higher number meaning a more powerful graphics card. Nvidia does sometimes add extra letters or words at the end of its graphics cards, with ‘Ti’ and ‘Super’ being common variants. This typically means a slightly more powerful card. For instance, the Nvidia GeForce RTX 5070 Ti is going to be slightly more powerful than the RTX 5070. The basic rule of thumb is that, after the number, any extra letters or words typically indicate a faster graphics card.
When it comes to specs, graphics cards have a lot of numbers and lingo to swallow, which can definitely be overwhelming. But if you just want to slot a GPU in your rig and forget about it, you don’t need to pay attention to every little detail.
The amount of VRAM you need is going to largely depend on what resolution you want to play your games at. If you want to play games at 4K, you’re going to want as much video memory as you can afford. There are games that will eat up upwards of 20GB of VRAM at that resolution if you let them, so the sky really is the limit. For lower resolutions like 1080p, however, you can get by with 8GB of RAM, though I would recommend going with a 12GB or 16GB card if it’s in your budget. An 8GB card will get you through most games, but there is a growing number of AAA PC games that need more memory.
There are some other specs you can look at if you really want to, like clock speed, CUDA core count and Streaming Multiprocessors – what Nvidia calls its compute units – but those largely just get better as you get a more high-end graphics card either way. Compute Units are essentially the same as CPU cores on a processor, and each one contains 128 CUDA cores. So, the RTX 5080 with its 84 Compute Units, contains a total of 10,752 CUDA cores. Just keep in mind that directly comparing two graphics cards of different generations by the core counts alone won’t tell you how much of a performance jump it’ll be.
Once you’ve picked the graphics card you want, you need to make sure you have a power supply that can handle it. You can usually check the box your graphics card comes in to get an idea of how much power the card requires, and if not, Nvidia has it listed out on its website. My advice would be to check the required power for your graphics cards and then get a power supply that can provide just a bit more wattage. For instance, if your graphics card recommends a 450W power supply, I’d get a 550 or a 650W unit just to be on the safe side.
1. If You Just Want the Best: RTX 5090
The Nvidia GeForce RTX 5090 is the most powerful graphics card on the market right now, there's no way around that. But I still wouldn't recommend most people buy it. This is a $1,999 graphics card to start and it's not exactly super available on store shelves right now. We've even started seeing scalpers selling this thing for $9000 on eBay, just a day after launch. If you do have the funds to get your hands on it, though, you can expect the best 4K gaming experience money can buy, even if it isn't light-years ahead of the RTX 4090.
When I reviewed the Nvidia RTX 5090, I found that it's around 20-25% faster than the RTX 4090 at 4K, with the performance lead obviously waning at lower resolutions. That's a relatively small gen-on-gen performance uplift, but at the end of the day, faster is faster. The RTX 5090 largely gets its larger performance from an absolutely massive GPU, with 21,760 CUDA cores, which can boost up to 2.41GHz. That's a sizable jump from the RTX 4090, and it's possible primarily through a much larger power budget. This graphics card requires 575W of power to run, which is the most power a consumer graphics card has ever required – including back in the days of dual-GPU graphics cards like the AMD Radeon R9 295X2.
As such, you're going to need a serious power supply to keep this GPU fed with power. Nvidia recommends at least a 1,000W unit to pair with the RTX 5090, but I'd recommend going a little beyond that, with something like a 1,200W PSU, just to make sure it stays efficient when you're really pushing it to its limits. Because this graphics card will hit that power limit, especially when DLSS 4 Multi-Frame Generation is enabled, and that's largely the reason to get this graphics card.
Essentially, DLSS 4 Multi-Frame Generation, or MFG, uses the AI Tensor Cores to generate up to 3 AI frames off of each rendered frame. This can hugely improve your framerate, but can introduce serious latency if you're not already getting a solid frame rate. This is a feature you should only really enable if you're already getting 60-70 fps, but it's going to be rare to find a PC game that the 5090 is going to have any trouble hitting that frame rate on.
2. Best for 4K (for Most People): RTX 5080
Every GPU generation is a gamble of whether or not it's going to be a huge improvement on what came before, or just fall flat. The Nvidia GeForce RTX 5080 kind of leans towards the latter option, but that's not to say it's a bad graphics card. While I wouldn't recommend anyone who has an RTX 4000 graphics card to upgrade to the 5080, it's a good upgrade for anyone that's been waiting a couple of generations for a new graphics card.
In my review, the RTX 5080 only ended up being around 11% faster and 8% faster than the RTX 4080 and RTX 4080 Super, respectively, at 4K. That's one of the weakest generational uplifts for a graphics card in years, and it remains slower than the last-generation flagship, the RTX 4090. But given that the RTX 4090 is still way more expensive than the RTX 5080, this is still the best Nvidia graphics card you're going to get around a thousand bucks – assuming you can find one.
The main reason the RTX 5080 isn't much faster than its last-gen counterpart is because Nvidia didn't shrink the manufacturing process, remaining on a similar 4nm node, while also not drastically increasing the amount of CUDA cores on offer. Luckily, that helps avoid the massive power requirements seen on the RTX 5090, with the 5080 'only' requiring 360W of power to run. That means you don't need to fork over the cash for an expensive high-wattage power supply, which means more cash for the games you want to play.
And while this graphics card might be a little disappointing for anyone who keeps a laser focus on graphics cards every generation, playing games on this thing is awesome. Throughout my entire test suite, the only game that didn't get 60 fps at 4K was Metro Exodus: Enhanced Edition, and that was because I was running it without DLSS with Ray Tracing cranked up. Turn on DLSS, and that number is going to hit the triple digits, just like pretty much any AAA game on the market right now. Add in DLSS 4 Multi-Frame Generation, and you should have no problem fully saturating a high-refresh 4K monitor. Just make sure you can hit 60-70 fps before you turn it on.
3. Best for 1440p: RTX 4070 Super
If you ask me, 1440p is the gold standard PC gaming resolution. Not only are the monitors much cheaper than their 4K brethren, but they’re much easier to power with affordable graphics cards. You can absolutely strap an RTX 4080 Super into a gaming PC to play games at 1440p, too, but you’re better off dialing in the performance and saving more money for games. The Nvidia GeForce RTX 4070 Super is the perfect Nvidia graphics card for 1440p, no matter what games you’re playing.
In my review, I found that the RTX 4070 Super is easily able to play the most demanding games like Cyberpunk 2077 at 1440p, delivering 91 fps on the Ray Tracing Ultra preset. And in games like Forza Horizon 5, that number goes up to 158 fps, proving the RTX 4070 Super is capable of high frame rate gaming at 1440p. And at $599, it’s much more affordable than the RTX 4070 Ti or RTX 4070 Ti Super, which will set you back $749 and $799, respectively.
The only downside is that Nvidia didn’t upgrade the VRAM when refreshing the RTX 4070. You’re still getting 12GB of GDDR6X memory, which should be plenty for most games, but you might find it getting stretched in more demanding games like Black Myth Wukong. This is especially unfortunate given you can get the RTX 4060 Ti with 16GB of VRAM at a lower price – even if the RTX 4070 Super will outperform the 4060 Ti in every game regardless.
4. Best for 1080p: RTX 4060 Ti
According to the latest Steam Hardware Survey, 1080p is still far and away the most popular display resolution for PC gamers. While that number has been going down over time, it makes sense why so many people still play at 1080p. It’s simply more affordable to play at this resolution, as you don’t need an extremely powerful graphics card to play even the most impressive AAA games. Plus, because of the low power demands for this resolution, it remains extremely popular among esports players, who can get insane frame rates that are just not possible at 4K.
And that’s what makes the Nvidia GeForce RTX 4060 Ti such a great 1080p graphics card. The RTX 4060 Ti is affordable at $399, and is easily able to play any game at 1080p at a high frame rate – and I don’t mean just 60 fps. You see, in my review, I found the RTX 4060 Ti to exceed 100 fps in most of the games I tested. In fact, the only game that it didn’t top 100 fps in was Cyberpunk 2077, with 76 fps, but that number went all the way up to 122 fps once I enabled Frame Generation.
In most of the games where a super high frame rate will actually matter, the RTX 4060 Ti can easily reach upwards of 200 fps. Games like Overwatch 2, Valorant and Counter-Strike 2 will love the extra frames this GPU can spit out.
One thing you have to be aware of, though, is there are technically two different versions of the RTX 4060 Ti: An 8GB model for $399 and a 16GB version that costs $499. For most people who just want to play games at 1080p, the 8GB version is going to be fine. And while, yeah, 16GB is better, it’s really not worth adding an extra $100 to the price tag for the two games that need more than 8GB at 1080p.
5. Best on a Budget: GeForce GTX 1660 Super
If $399 is still too expensive, you can still pick up the Nvidia GTX 1660 Super even if it's an ancient graphics card by today’s standards. The GTX 1660 Super is built on the Nvidia Turing architecture that powered the RTX 2080 back in 2018. However, while it’s more than a little outdated at this point, it can still deliver solid 1080p gaming, especially in less demanding games like League of Legends.
If you’re going to compromise and get this older graphics card, keep in mind that you’re missing out on many of Nvidia’s best features. This GPU doesn’t have RT cores, so it doesn’t support ray tracing, and it doesn’t have Tensor cores, which means no DLSS.
When I reviewed the GTX 1660 Super for TechRadar, I was amazed by just how well it performed at 1080p for the price. That was five years ago at this point, but Nvidia still hasn’t released a direct successor to this budget graphics card. That’s a shame, because Nvidia’s graphics architecture has grown a lot since then, and another budget card is long overdue. Luckily, Nvidia still hasn’t discontinued the GTX 1660 Super in the years since it launched and you can find it for as little as $180. Five years later and the Nvidia GeForce GTX 1660 Super is still the best Nvidia graphics card under $200, however shameful that is.
Upcoming Nvidia Graphics Cards
The first wave of Nvidia Blackwell graphics cards are all here, and it seems like that's all we're going to get for now. Usually, Team Green would follow up its first wave of graphics cards with refreshed "Super" variants. However, nothing in the world of PC gaming hardware is normal right now, and it looks like the Super lineup has been canceled in the face of high RAM prices and AI datacenter demand. Right now, it looks like we'll be waiting for the RTX 6090 whenever that comes out – likely sometime in the second half of 2027.
What Is DLSS?
Deep Learning Super Sampling, or DLSS is an AI upscaling method that uses Nvidia’s Tensor Cores to improve image quality. The Tensor Core will take visual data from the frame being rendered, along with motion vector data, in order to accurately upscale the image to a higher resolution. This process results in an image that looks pretty close to the native resolution, but with a much higher frame rate.
In its early days, DLSS needed to train the AI model on each game it would support, with developers needing to upload data to Nvidia. However, DLSS has gone through several iterations, and now it does not need to be trained on individual games. Instead, if a developer wants to include DLSS in their game, they can just inject Nvidia’s API into the game. This means more games can support the technology, without having to wait for Nvidia’s training on each game.
With DLSS 3.0, Nvidia added Frame Generation to the equation. This technology takes the visual data from two frames, along with motion vector data from the game engine and motion information from its own hardware, and creates an entirely new frame that’s sandwiched between the two original frames. This would introduce a lot of latency, but Nvidia requires its Reflex technology to be enabled before Frame Generation can happen. Reflex essentially syncs the graphics card and the processor, eliminating the need for the CPU to queue up frames for the GPU to render later.
Nvidia updated to DLSS 4.0, then later 4.5, which introduced multi-frame generation and a new upscaling model. The new frame generation lets RTX 5000 graphics cards use AI to generate up to three frames per rendered frame, greatly improving framerates. The new DLSS also moves the upscaling algorithm to a transformer model, which allows for much better image quality, though with a slight hit to performance. But, hey – that's what frame generation is for, I guess.
Together, Frame Generation and Reflex greatly improve your frame rate, but there’s a catch. Because so much of it relies on motion data, you already need to have a decent frame rate for Frame Generation to work smoothly. So, this technology is best used for folks that can already get 60fps or more and just want to push to a higher frame rate.
What Is Ray Tracing?
Ray tracing is just a way to render light realistically. It does this by taking a light source, then simulating each ray of light as it bounces around the scene. It’s a simple concept to be sure, but it ends up requiring a ton of compute power to pull off. Any light being cast potentially has thousands of rays of light, each of which will bounce around hundreds of times, multiply that by needing a new frame 60 times a second, and you can imagine how much power you need to pull it off.
That’s why ray tracing in video games needs specialized hardware to pull this lighting method off without grinding your framerate to a standstill. Luckily, Nvidia has been working this hardware into its graphics cards since the RTX 2080 in 2018.
Fast forward to today, and even with that dedicated ray tracing hardware being built into every mainstream graphics card – not just from Nvidia – we still need to limit ray tracing in order to maximize performance. There are only a few games that support full ray tracing, or ‘path tracing,’ with most other games limiting the ray tracing elements to a certain part of the scene, like shadows or reflections, and also limiting the amount of bounces calculated for each ray of light.
It’s an expensive way to generate lighting, but it looks incredible, especially in games that rely on lighting for atmosphere. In Metro Exodus, for instance, the accurate lighting often leads to darker environments, amplifying the intense atmosphere the game already has.
Jackie Thomas is the Hardware and Buying Guides Editor at IGN and the PC components queen. You can follow her @Jackiecobra
Connexion
