The Best Graphics Cards for 2022

Graphics Card Basics: Understanding the Core Specs

Our comparison charts above should help you narrow down your options based on your monitor’s resolution and the card families you’re interested in (and your budget). The graphics processor’s clock speed, onboard VRAM (i.e. how much video memory it has), and—of course!—price are all important considerations when comparing cards, though.

Clock Speed

High base clock speeds (the speed at which the graphics core operates) and more cores are indicative of a faster GPU when comparing identical GPUs from the same series. However, only cards from the same product family, using the same GPU, can be compared in this way. The Founders Edition GeForce GTX 3080 has a base frequency of 1,710MHz, while MSI’s Gaming X Trio version of the RTX 3080 (using the same chip) has a base speed of 1,815MHz in its out-of-the-box Gaming mode.

The graphics chip’s “boost clock” is not included in this base clock measurement. The graphics chip’s boost clock allows it to momentarily increase its speed when it is under load and thermal conditions permit it to do so. This can also differ from one card in the same family to the next in terms of design. A card’s cooling hardware and factory settings are two factors that can affect how well it performs. Boost clocks for a particular GPU tend to be higher on high-end partner cards with multi-fan coolers.

That’s not even including AMD’s own proprietary GPU speed category: “game clock.” Gamers should expect to witness a “average clock speed across a wide range of titles,” according to the company’s engineers, who tested 25 different games on the company’s RDNA and RDNA 2-based portfolio of cards. So that you don’t compare game clocks to boost or base clocks, which game clock clearly isn’t.

Understanding Onboard Video-Card Memory

Onboard video memory (also referred to by the old-school name “frame buffer”) is usually suited to the needs of games or programmes that the card is built to execute. As long as you’re playing PC games at the appropriate resolutions and detail levels, you may assume that your video card will have enough RAM to keep up. If a card manufacturer overprovisions a card with more memory than it can realistically consume, the price and competitiveness of the card will suffer. There are, however, some kinks in this plan.

Generally speaking, a card built for 1080p gaming these days has 4GB or 6GB of RAM, whereas cards designed for 1440p or 4K gaming have 8GB or more of RAM. For the most part, the amount of RAM on a specific GPU-based card is the same across the board.

There are a few wrinkles: Card manufacturers offer variations of a card with the same GPU but varying quantities of VRAM in some isolated but crucial circumstances. Cards based on the Radeon RX 5500 XT, for example, are available in 4GB and 8GB configurations. So keep an eye on how much RAM both cards have because they’re both midrange GPUs. The less expensive ones will offer fewer features.

A video card with at least 4GB of memory should be mandatory if you’re searching for an all-out 1080p experience. AMD and Nvidia now include more VRAM than this in their $200-plus MSRP GPUs. While Nvidia’s GeForce RTX 3090 has 24GB of GDDR5 memory, AMD has 8GB on its RX-series cards, while the latter has 16GB on its top-of-the-line cards.) Any card with less than 4GB of RAM should only be used for low-resolution gaming, basic or outdated games that don’t require a lot of hardware resources on the part of the user.

It’s a completely other ballgame for creators. The amount of onboard VRAM is more significant in many 3D rendering programmes (as well as VFX operations, modelling, and video editing) than the boost clock speed. In most circumstances, the better a graphics card is for rendering out a complicated VFX picture with thousands, if not millions, of distinct elements, the more VRAM, faster memory, and a greater bandwidth pipe it has.

This is only one of the many specifications you’ll come across. It’s a measure of how fast data can flow into and out of the graphics processing unit (GPU). Since AMD and Nvidia have different designs and, at times, demand different amounts of memory bandwidth, these figures cannot be directly compared.

The type of memory in your next GPU is also a vital consideration, and understanding which type you’re purchasing into is critical depending on the types of games or programmes you want to run, even if you won’t really have a choice within a card line, is crucial…………………………….

Even though AMD’s Radeon VII can theoretically keep up with an RTX 2080 when it comes to gaming, it also has 16GB of HBM2 VRAM for content creators to take advantage of. When it comes to video editing in Adobe Premiere Pro, HBM2 is favoured since it is cheaper to manufacture than its predecessor, HBM1. GDDR5 has since overtaken it in popularity.

Modern GPUs have relied on GDDR6 memory for decades, from the RTX 2080 Ti and RTX Titan all the way down to the AMD Radeon RX 5600 XT. If you’re looking for a VRAM solution that can run even the most demanding AAA games at a high resolution at a reasonable price, GDDR6 is an excellent choice. Radeon RX 6000 Series GPUs from AMD use GDDR6, which would seem like a good upgrade if Nvidia’s most recent cards didn’t already use…

Nvidia has begun using GDDR6X memory in its GeForce RTX 30 Series graphics cards. It more than doubles the original GDDR6 design’s possible bandwidth without encountering the signal deterioration or route interference issues that plagued prior incarnations.

Upgrading a Pre-Built Desktop With a New Graphics Card

For the most part, pre-built computers these days have the cooling capacity to accommodate a new discrete GPU without any issues.

Before purchasing or upgrading a graphics card, measure the available space in your chassis. It’s not uncommon for a gap to exist between the motherboard’s far right edge and the hard drive bays. In some cases, you may only have a few millimetres to spare on your GPU’s total length. In rare circumstances, extremely long cards can be a concern. Our top picks for tiny PC graphics cards may be found here.)

Next, measure the height of your graphics card. In some cases, the card partners have their own coolers that deviate from the normal AMD and Nvidia reference designs. The card you choose should not be too tall to close your case if it has an ornate cooler design.

Finally, we come to the power supply unit (PSU). You’ll need a power supply capable of supporting a new graphics card in your system. If you’re upgrading a pre-built PC that came with a low-end video card or no card at all to a high-end video card, be very cautious about this. The odds are against you if you’re looking at a low-cost or business-oriented PC, as their power supplies tend to be underpowered or underequipped.

Aside from the maximum wattage of your power supply, the number of six-pin and eight-pin connectors on your PSU is also an important consideration. A six-pin power connector for a video card is a common feature in most modern systems, including those sold by OEMs like Dell, HP, and Lenovo, and some have both a six-pin and an eight-pin connector.

A six-pin cable, an eight-pin cable, or a mix of the two is required to power midrange and high-end graphics cards. (Except for the lowest-end cards, which rely entirely on the PCI Express slot for power.) Make sure you know what kind of connectors your card requires before you buy it.

Since some of the GeForce RTX cards require a special adapter (it comes in the box) to turn two eight-pin PSU connectors into a single 12-pin one card-side, and MSI’s 12.7-inch GeForce RTX 3080 Gaming X Trio (and a few other high-end monsters) now require three eight-pin connector suck down its required juice, we’ve noticed some changes here. A third PSU connector isn’t required by just this one new RTX card; there are others as well.

The recommended power supply wattage for each of Nvidia and AMD’s graphics-card families is listed on their respective websites. Take these principles carefully, but keep in mind that they are merely guidelines and are generally conservative in their approach. Do not use a 300-watt power supply if AMD or Nvidia states you need at least a 500-watt power supply to run a specific GPU, but also know that you do not need an 800-watt power supply to guarantee ample headroom. You should always check the power-supply suggestion specifications for the exact card you are looking at before purchasing a third-party version of a given GPU.

SLI, CrossFireX, and NVLink: Fading for Gamers

Both AMD and Nvidia have been moving away from supporting dual-, tri-, or even quad-card systems over the last few generations. These have traditionally been a pricey, but rather simple, approach to boost performance. However, the value proposition is no longer there (especially for PC gamers).

With the addition of two RTX 2080 Ti cards to our testing in 2019, we discovered mixed results to put it bluntly. Today’s games rarely make use of several GPUs and those that do don’t show speed improvements as a result. Nvidia’s SLI and NVLink are twin-card technologies; AMD’s CrossFireX is a twin-card technology. There are several games that run slower because of engine optimization.

It’s a different scenario, however, when it comes to content development. NVLink card pairing is only useful for professional creators, which is why Nvidia’s GeForce RTX 3090 is now the only card in its current portfolio that supports it. It’s a good investment for them.

To sum it up? It is usually always better to go with the most expensive card you can afford rather than a less expensive one that you want to add to your collection in the future.

Ports and Preferences: What Connections Should My Graphics Card Have?

DVI, HDMI, and DisplayPort are the three most prevalent types of ports found on the back of today’s video cards. DVI is still used in some systems and monitors, but it is the oldest of the three standards and no longer appears on high-end graphics cards today.

There are usually three DisplayPorts (sometimes four) and one HDMI port on most graphics cards. There are various distinctions between HDMI and DisplayPort. The first requirement is that your graphics card must support HDMI 2.0a or DisplayPort 1.2/1.2a at the very least if you intend to use a 4K display today or in the near future. Even if the GPU supports HDMI 2.0b or DisplayPort 1.4, smooth 4K playback or gaming will still be a challenge without it. (The most recent generation of cards from both manufacturers will suffice.)

For all of Nvidia’s GeForce RTX 30 Series cards, HDMI 2.0 has bandwidth restrictions of 18Gbps; the latest HDMI 2.1 cable specification has bandwidth limits of 48Gbps (in HDMI 2.1). 8K resolution can be displayed at a refresh rate of up to 60Hz, while 4K is supported up to a maximum of 120Hz.

The 20-series of GeForce RTX Turing cards from Nvidia use a connector called VirtualLink. DisplayPort over USB-C is also supported by this interface, which appears and functions like a USB Type-C connection. Virtual-reality (VR) headsets are what the port was really built for, offering enough power and bandwidth to meet the demands of VR head-mounted displays (HMDs). In theory, it’s a wonderful feature that isn’t available on the Founders Edition cards of the RTX 30 series, therefore its future is uncertain.

Looking Forward: Graphics Card Trends

This is the first time that Nvidia and AMD have engaged in a head-to-head battle over consumer video cards in more than a year, stirring things up between the two major players.

Nvidia DLSS, Super Resolution, and Image Sharpening Technology
The inclusion of image-sharpening technologies like as AMD’s Radeon Image Sharpening (RIS), FidelityFX with CAS (also from AMD), and Nvidia’s Freestyle has had a significant impact on the gaming landscape in the last year or two. The question is, what are these programmes, and how can they benefit gamers who are on a tight budget while purchasing games?

Render scaling is to blame for everything. Your graphics options may have an option to modify the size of the game’s rendered area, which is common in most recent games. You may think of it as reducing the “render” resolution to something like 2,048 by 1,152 pixels while still maintaining the existing game resolution (let’s say 2,560 by 1,440 pixels) (again, for the sake of this example).

Then again, who would intentionally make their game appear bad? Those that use game sharpeners are the ones to blame. Scaling down a game’s render resolution using image-sharpening technologies increases frame rate (since lower resolutions use less GPU muscle to draw), while an image-sharpening technology cleans things up in the background at a minimal performance cost.

A sharpening filter is used to restore detail to the downsampled image as part of the “cleaning things up” process. You can theoretically gain a large amount of performance without sacrificing much visual clarity if you set it just right (85% down-render with a 35% sharpen scale is a popular ratio). What’s the significance of this? If you’re able to reduce the frame rate of your game without sacrificing visual quality, you’ll be able to get by with a cheaper video card.

During our testing, we determined that the peak down-sample is roughly 30%. For the same high-definition gaming experience as running the game at its native resolution without render scaling, you can buy a card that’s nearly a third cheaper than what you were originally looking at, sharpen it back up to 30% using one of the aforementioned sharpening tools and still get close to the same experience.

With regard to smoothing out polygons around the edges of characters and objects, Nvidia has come up with DLSS, or “deep learning supersampling,” which stands for “deep-learning supersampling.” Video game anti-aliasing is one of the most computationally intensive tasks for a graphics card to perform in video games, and it has been employed from its creation to achieve the same goal: to make the jagged thing look smoother.

Nvidia makes use of artificial intelligence to assist in DLSS. Although it’s expensive (i.e., requires a GeForce RTX card) for the time being, direct light scattering is more expensive (i.e., requires an RTX card) because it can’t be done on a regular CUDA core but rather a specialist graphics core called a Tensor core. The Tensor core is the ray-tracing equivalent of the RT core for Nvidia’s artificial intelligence neural network’s complex equations.

Because so few games support DLSS and its follow-up technology DLSS 2.0, there is a lot of room for improvement. With the addition of DLSS 2.0 support to Unreal Engine 4, this is beginning to change, although it will still be some time before it is widely used by game developers across all platforms and platforms. PC Labs observed that one of the most well-known DLSS-capable games, Death Stranding, benefited greatly from the utilisation of both DLSS and CAS during our testing. The visual quality was actually better with DLSS than CAS, despite the fact that there were still some noticeable twitches while characters were moving. GeForce RTX card users could benefit greatly from more games adopting DLSS. A significant “if” comes into play here.

Also in this bout in 2021, AMD threw some punches with its FidelityFX Super Resolution technology. More than 30 games have been added to the compatible titles list since we tested the functionality way back in June 2021, despite the fact that only a few supported this upsampling technology when it launched in mid-2021. Nvidia and AMD’s latest video cards are supported, allowing for a far wider range of compatibility than before. DLSS, on the other hand, requires a GeForce RTX graphics card to work.

In the first half of 2022, AMD launched its next upscaling weapon: Radeon Super Resolution. FidelityFX’s Super Resolution is based on the same principles and algorithms, but they are now implemented at the driver level. This restricts the GPUs it operates on—only AMD Radeon RX 5000 and RX 6000 models—but opens up tens of thousands of games to utilise the feature. More information on the technology can be found by clicking here.

VR: New Interfaces, New HMDs?

There is a lot of space for improvement because only a few games support DLSS and its follow-up technology DLSS 2.0. Things are changing, but it will be some time until game developers across all platforms and platforms start using DLSS 2.0 support in Unreal Engine 4. PC Labs found that Death Stranding, one of the most well-known DLSS-capable games, benefited considerably from both DLSS and CAS during our testing. Despite the fact that there were still apparent twitches while characters were moving, the visual quality was actually better with DLSS than CAS. Users of the GeForce RTX card could tremendously benefit from additional games incorporating DLSS technology. There’s a big “if” in this situation.

The FidelityFX Super Resolution technology from AMD was also used in this fight in 2021. Despite the fact that only a few games supported this upsampling technique when it premiered in mid-2021, more than 30 games have since been added to the list of compatible titles. Allowing for a much broader range of compatibility than ever before, Nvidia and AMD’s most recent video cards are supported When it comes to DLSS, you’ll need a GeForce RTX graphics card in order to get it to operate.

Next year, AMD unveiled its Radeon Super Resolution, a new tool for upscaling images. The same principles and algorithms underlie FidelityFX’s Super Resolution, which is applied at the driver level. Although this feature can only be used on AMD Radeon RX 5000 and RX 6000 graphics cards, it frees up the feature for thousands of new titles. You can learn more about the technology by clicking this link.

However, two new headsets released in 2019 increased the power consumption slightly. There are now 2,560 by 1,440 pixels per eye in the Oculus Rift S and 1,440 by 1,600 pixels in the $1,000 Valve Index, bringing the total number of pixels to 2,880 by 3,200 pixels. As a result, the newest generation of top-tier GPUs from Nvidia and AMD, which were released in 2021 and 2022, both chose not to include a virtual link connection on the card’s backside.

You’ll need a graphics card that can keep up with the rigorous demands of these newest headsets if you opt to spend the money (80Hz refresh on the Rift S, and 144Hz refresh on the Index). This indicates a machine that can run Half-Life: Alyx at 144fps on two monitors simultaneously, above 1080p, which is a graphically intensive game. To get the greatest experience, Valve recommends at least an RTX 2070 Super or an AMD Radeon RX 5700 XT. To summarise, be sure to read the specifications of any headset you are contemplating and strictly adhere to the GPU recommendations. A headache and even nausea are the only things you’ll get from a subpar VR experience.

So, Which Graphics Card Should I Buy?

As the year 2022 begins, and GPUs are still in short supply, the answer to this question is more muddled than ever. When it comes to games and graphics cards, new technology is making things more complicated for both the high-end and mid- and low-end gamers, because of image-sharpening factors (and possible discounts).

RTX 3080, RTX 3080 Ti, and RTX 3090 from Nvidia are now the only cards that can match their MSRP price-to-performance ratio when it comes to the high end of the market. If you’re searching for maximum power, Nvidia is your best bet.

As for Nvidia, the company’s new Radeon RX 6000 Series is bringing AMD back into the high-end market, although it’s not quite as competitive as Nvidia’s current offerings. Some games still have driver implementation issues with the RX 5000 line one year after the RX 5700 and RX 5700XT were released. This could be a red flag for certain customers.

Lower down the line, Nvidia and AMD butt heads regularly, each with their own answers to sub-$1,000 PC builders that span a huge number of different card models and available options.

The GPUs we recommend at the top and bottom of this article span the spectrum of budget to high end, representing the full range of the best cards that are “available now.” Note, though: Those quote marks are very intentional.

The Elephant in the Room: Graphics Card Pricing (and Availability) in 2022

Because of a supply-side difficulty that began in 2020 and continued into 2022, discrete graphics cards are only now becoming widely available at all price points, from enterprise to consumer.

This is all you need to know, condensed: First and foremost, it is not the “fault” of any one entity. Not AMD’s, not Nvidia’s, and certainly not Bitcoin’s (on its own). Several elements have come together to create the pricing situation that we are familiar with at the time of this writing in early 2022. Even before the outbreak of the pandemic, experts in the field predicted that we would be heading in this direction.

Every industry that relies on semiconductors, from graphics cards to the automotive industry, is currently experiencing a supply pressure. It’s possible that the situation is worse in GPUs. While the card scarcity in 2017-2018 was caused by the initial rise in bitcoin prices, this time around, cryptocurrency is growing at the same time that tariffs have hit, on top of an increase in bots meant to get over Newegg and Amazon’s shopping-cart captcha limits. Card sales are now limited to just a few minutes after they go on sale due to this phenomenon. Once they’ve been stripped of their original packaging, they’re sold for a significant markup on sites like eBay and elsewhere. This is a price that some purchasers in the market are definitely ready to pay.

In the long term, many in the sector expect this particular squeeze to continue. Furthermore, as long as supply bottlenecks persist, demand for discrete desktop GPUs will rise. As demand for last-generation GPUs declines ahead of the arrival of the new generation, prices will continue to rise.

In the past year, we’ve seen GPUs like the GTX 1080 Ti sell for $100 more on eBay than they did at MSRP when they were first released.

GPUs on eBay at Different Prices
Folks, it’s a wild place out there.
The point is that none of the pricing we’ve examined so far match the current reality for purchasers at the MSRP level. Use this guide as a benchmark for comparing the relative performance of the card you’re considering to those of the competition. But what about in the actual world? Depending on what time of day you browse Best Buy, Newegg, eBay, or Amazon listings, you can come across a winning lottery ticket or a loss. In today’s market, there are no assurances. Anybody’s guess as to whether or not you’ll receive what you want and at what price. The best you can hope for is to get a card for less than the MSRP using our methods. It is doable, but it is not simple.

As a result, “Any cards you can find at a decent price” are, to some extent, “The Best Graphics Cards for 2022.” If you’re intending to wait until the next generation of GPUs comes out, the Nvidia GeForce RTX 4090 will be available on April 1st, which should help with supply.