Video cards tests: PassMark Software — Video Card (GPU) Benchmark Charts

How We Test Graphics Cards

Video cards. Graphics cards. GPUs. Whatever you call them, there’s no denying the important role these components play for gamers, not to mention workstation users and productivity-focused folks with multiple displays. Whenever you’re operating a PC to render real-time graphics, a dedicated graphics card will prove itself useful. But as you’re shopping around for video cards, it is crucial to consider the performance you get for the amount of money you’re shelling out. That’s where our testing comes in.

Here at PCMag, we run the numbers on every graphics card we get our hands on to determine which models are the most capable. With so many graphics cards to choose from, we have to ensure that our testing regimen remains consistent with each one that we review. For that reason, we use a variety of standardized benchmarks to test video cards, some of which were developed for the sole purpose of pushing discrete graphics components to their limits. Other benchmarking tools are found inside of leading, AAA-grade video game titles you may recognize, put into place by their respective developers.

Our 2020 Graphics Card Testbed

All graphics cards that come through the PCMag offices are tested on a custom Windows 10 desktop PC we built expressly for the purpose of putting GPUs through their paces. The 2020 version of this machine is equipped with an Asus ROG Maximus XII Hero Wi-Fi(Opens in a new window) motherboard, an Intel Core i9-10900K CPU, 16GB of G.Skill TridentZ RAM, a 2TB Samsung SSD 870 QVO SSD, and a Thermaltake Toughpower Grand 1,050-watt power supply (PSU).

We use this testbed to run a variety of synthetic tests (which simulate graphical stresses or game processes), as well as actual in-game benchmarks that report results in frame rates or game-specific scores. Windows 10 is always patched to the most recent version, and we use the latest compatible version of the graphics driver recommended by the card maker for the card being tested.

Synthetic Benchmarks

UL 3DMark

Our core selection of synthetic benchmarking tools is headlined by the seminal 3DMark(Opens in a new window), branded by its developer as «The Gamer’s Benchmark.» Developed by Underwriters’ Labs (UL), 3DMark is host to a suite of graphics subtests designed for different classes of graphics cards.

The 3DMark Fire Strike Ultra subtest is the first test we run when testing graphics cards at PCMag, and it’s meant for stressing graphics muscle at a simulated resolution of 4K (3,840 by 2,160 pixels). Fire Strike Ultra doesn’t require a 4K monitor to run, but it does call for a graphics card with at least 3GB of video memory (VRAM).

The next 3DMark subtest we run, Time Spy Extreme, is a DirectX 12 benchmark, also simulating the 4K resolution of 3,840 by 2,160 pixels (aka 2160p). Time Spy Extreme requires a graphics card with at least 4GB of VRAM.

The last 3DMark test we run, Port Royal, is more of a specialized performance benchmark. We use for a specific purpose; more about it below in our discussion of ray-tracing and Nvidia DLSS.

At the end of each 3DMark test, we are given an Overall Score and a Graphics Subscore, each usually somewhere in the thousands when testing discrete graphics cards. We report the Graphics Subscore, which isolates the video card in terms of its contribution to the task. The higher the score, the better the card’s graphical performance relative to others. The scores generated by 3DMark are useful only when compared with other 3DMark scores from other hardware tested until identical conditions.

V-Ray & LuxMark

V-Ray is a simple test, but a necessary one in the era of modern GPUs. V-Ray calculates how well a card can handle path tracing, a lighting technique commonly used in 3D rendering and animation. The number we see at the end is the number of millions of paths (mpaths) that were rendered during the one-minute benchmark.

LuxMark, meanwhile, is an oldie but a goodie. It tests how well a GPU can handle OpenCL-based engines. OpenCL is an aging standard, but it is still considered a «standard» for a reason. LuxMark is measured as a value in a vacuum: The only thing you can compare a LuxMark score against is itself!

FurMark

FurMark is another aging benchmark that has stuck around for a reason: For many successive eras of cards, it has held the job of stress-testing those GPUs to their limits.

The FurMark workload is more of a test of what a card can take rather than what it can do, giving you a good idea of how a GPU will hold up with fully loaded rendering queues over an extended period of time. We run FurMark at the three resolutions as indicated in our performance charts (1,920 by 1,080 pixels; 2,560 by 1,440 pixels; and 3,840 by 2,160 pixels) and report the scores from each.

Unigine Superposition

After the various 3DMark runs, the next benchmark we execute is Unigine Superposition(Opens in a new window), which uses Unigine’s own graphics engine. It is a demanding stability test complete with its own backstory. It takes place in a deserted classroom; we see renderings of a professor’s scientific instruments and inventions, leading into an animated segment where the room’s contents begin floating in the air, magnetized to a mysterious machine.

The Superposition testing utility comes with a variety of presets, which each govern both the resolution the test is run at and the level of detail. We use three of these: 1080p Extreme, 4K Optimized, and 8K Optimized. Superposition generates a points-based score after the conclusion of each sequence, along with minimum, maximum, and average frame rates. We record and report the average frame rate at each of the three presets we use.

After Superposition is complete, we move on to a long sequence of games accessed via Steam, Uplay, or the Epic Games launcher, depending on the title. Let’s get into those next.

AAA Games

All of our AAA game tests are reported in average frames per second, except where indicated. Most employ a built-in benchmark sequence provided by the game developer, but in a few cases where indicated, we have developed a custom run.

Shadow of the Tomb Raider

Though the game released to mixed reception when it came out in September 2018, critics all agreed that Shadow of the Tomb Raider(Opens in a new window) is a visual stunner. On PC, it’s pretty demanding, too, with publisher Square Enix recommending at least an Nvidia GeForce GTX 1060 to achieve playable frame rates at the highest settings. And when we say «highest,» we mean it, because instead of calling it Very High or Ultra, the highest preset graphics setting in Shadow of the Tomb Raider is literally called «Highest.»

After configuring Shadow of the Tomb Raider to run at the Highest setting, we make sure DirectX 12 is enabled, and then we run the in-game benchmark (found in the Display and Graphics menu) at three full-screen resolutions: 1,920 by 1,080 (FHD; 1080p), 2,560 by 1,440 (QHD; 1440p), and 3,840 by 2,160 (UHD; 4K). Each time the benchmark completes, we record the average frame rate it gives us at the end.

Next, we run the game with both RTX (ray-tracing acceleration; more on that later) and DLSS turned on, where applicable. (This is card-dependent; only GeForce RTX cards support these features.) This gives us a good idea of how earlier implementations of DLSS (1.0, in the case of this game) will perform, and it also gives a baseline for how technology like ray tracing and DLSS are improving in newer titles like Death Stranding versus the original implementation.

Far Cry 5

The test we run next is Far Cry 5(Opens in a new window). An open-world first-person shooter from Ubisoft, Far Cry 5 uses DirectX 11 (DX11) for its Dunia Engine. It has a built-in benchmarking utility that shows a brief battle scene in an idyllic Montana lakefront landscape, representative of those in the game.

For newer cards that are optimized for DirectX 12, the results are often not as glamorous as you might expect in this benchmark. It’s also widely accepted that the Far Cry games tend to be optimized for AMD Radeon cards; thus, we’ve found that in past flagship cards, AMD has prevailed. The older Radeon RX Vega 64, for example, performs slightly better than the Nvidia GeForce GTX 1080 in Far Cry 5.

We use the top (Ultra) graphical preset and run this benchmark at 1,920 by 1,080, 2,560 by 1,440, and 3,840 by 2,160.

Final Fantasy XIV: Shadowbringers

Once we’ve recorded scores from Far Cry 5, we move along to the Final Fantasy XIV standalone benchmark. In this test, we run a benchmark for what is one of the most popular MMORPGs played online today. The game isn’t incredibly demanding, but it is a good indicator of what kind of performance MMORPG players as a whole can expect, given that this is one of the most graphically rich of the genre.

We use a custom variation of the High setting and run the test at 1,920 by 1,080, 2,560 by 1,440, and 3,840 by 2,160. When the Final Fantasy XIV test is finished running (and that does take a while, considering it presents one of the more extensive canned benchmark sequences), we are given a proprietary number that ranks the card among its peers. (It’s not a frames-per-second measure.)

F1 2020

F1 2020(Opens in a new window) represents a highly-tuned, modern game engine (no pun intended) that showcases one full three-lap F1 race, featuring dozens of high-fidelity cars and tracks being rendered in high detail. We run the test at the Ultra High detail setting using the game’s Australia racetrack, with weather and environmental settings at their defaults.

Like the preceding games, we run F1 2020 at the three major resolution tiers (1080p, 1440p, 2160p). The engine itself, despite its gorgeous visuals, isn’t extremely demanding, though, with cards as low on the totem pole as the GeForce RTX 2070 Super Founders Edition managing to keep the benchmark above 60fps at 4K with all settings maxed out.

Red Dead Redemption II

Welcome to the system killer of 2020. Though it doesn’t roll off the tongue quite as easily, «Can it run Red Dead Redemption II?» is the new «Can it run Crysis?» of modern AAA gaming. The game sports some of the most graphically demanding visuals ever attempted in the medium.

RDR2 is gorgeous, and the benchmark is thorough, to say the least. It runs protagonist Arthur through every ecosystem of the map, from the peaks of snowy mountains down to the gutters of Blackwater and the alleyways of Saint Denis…twice. We use the benchmark’s Ultra quality setting at our usual three tested resolutions (1080p, 1440p, 2160p).

Even the flagship of Nvidia’s new Ampere lineup, the GeForce RTX 3080 Founders Edition, only manages to get just above 65fps in 4K on this title at Ultra, which means it’ll be stressing out cards at (or below) the enthusiast tier for years to come.

Assassin’s Creed: Odyssey

The Assassin’s Creed franchise is a long-standing pillar of the gaming community at this point, and though opinions may differ on the Odyssey(Opens in a new window) entry in the series as an open-world title, its visual splendor is hard to deny.

This is a game that relies on huge, expansive vistas to give players a concept of where they are on the map, so cards that can handle longer draw distances tend to perform better in this run than others. We use the benchmark’s Ultra High quality preset at our three tested resolutions (1080p, 1440p, 2160p).

Like Red Dead Redemption II, AC: Odyssey is a game that pushes most cards to the max of what they’re capable of, and only one card tested thus far, the GeForce RTX 3080, has been able to maintain a steady rate of over 60fps at 4K resolution.

Death Stranding

Death Stranding(Opens in a new window) is one of just a few games that are currently running with the newest version of DLSS activated, and one of of even fewer still that is running it without any kind of ray-tracing tech applied underneath. Death Stranding is one of the best games to benchmark in 2020 because like F1 2020 and Shadow of the Tomb Raider, it has the option for both DLSS and AMD’s competing tech, FidelityFX with CAS, to be activated directly from the menu.

These disparate options share a similar goal: add as much performance to your game as possible, while sacrificing the least amount of visual fidelity possible. We send the cards through six runs in total at the Very High preset: three with temporal anti-aliasing (TAA) turned on (one at each resolution), and then three with either DLSS or CAS activated (CAS keeps TAA turned on, while DLSS does not), depending on the brand of card. The three runs are done at the usual 1080p, 1440p, and 2160p.

For example, a GeForce RTX 3080 Founders Edition would get its three runs with DLSS turned on and set to Performance mode, while a GeForce GTX 1660 Ti would be run with TAA and FidelityFX upscaling turned on at the same time. Again, all runs are set to the Very High preset, regardless of the sharpening or upscaling tech being applied on top. The performance can be directly compared, though DLSS will always come out looking like the better option thanks to its advanced integration of AI via the Tensor cores inside every RTX card.

Recommended by Our Editors

How We Test Laptops

How We Test Desktop PCs

How We Test SSDs

Total War II: Warhammer

Total War II: Warhammer(Opens in a new window) is a good example of how a card will handle games that have hundreds, even thousands, of individual elements onscreen at once.

We run the Skaven benchmark at the Ultra preset, which simulates thousands of soldiers from two warring factions clashing with one another across multiple battlefields. The camera swoops in and out of the action, providing a good amount of variance for a GPU to achieve large ranges of performance for the final average frame-rate result. The three runs are executed at the usual 1080p, 1440p, and 2160p resolution settings.

Metro: Exodus

Metro: Exodus(Opens in a new window) was one of the first games to feature DLSS 1.0, and like Shadow of the Tomb Raider, it remains a good example of how early iterations of that and ray-tracing tech will work on newer and older GeForce RTX cards alike. The benchmark is run at the RTX preset with RTX and DLSS turned on if we’re testing an RTX card, as well as at the Extreme preset if we’re testing either a GeForce GTX card or a Radeon card from AMD. (RTX cards get a test run at Extreme, too.) The six runs are done at the usual 1080p, 1440p, and 2160p.

The game continues to be one of the most demanding of our suite, both in the advanced lighting technology and the sheer visual quality of all the elements on screen. DLSS 2.0 has surpassed 1.0 in its effectiveness for finding performance seemingly from nowhere, but Metro remains a solid test with an engine that keeps even the most powerful cards crushed under a pile of polygons.

Multiplayer Titles

Counter-Strike: Global Offensive & Tom Clancy’s Rainbow Six: Siege

To test how well a video card will hold up in popular multiplayer arenas, we test two hyper-popular games: Counter-Strike: Global Offensive(Opens in a new window) (aka, CS:GO) and Tom Clancy’s Rainbow Six: Siege(Opens in a new window).

Though most of our above AAA game tests are maxed out in graphical fidelity to push the cards to their limit, multiplayer gaming is all about maintaining the best balance between graphical fidelity and frame rate. Because of that, we’ve kept the test sequences for those two titles tuned to more moderate settings for high-refresh-rate play.

On CS:GO, we run at the three usual test resolutions (1080p, 1440p, 2160p) with a preset setting of High, but MSAA set to 4x. On Rainbow Six, we use the High graphics setting as well.

Legacy Titles

We use a custom script in the now-deprecated (but still very useful!) Adrenaline Action Benchmarking Tool to run the in-game benchmarks in four legacy PC games: Tomb Raider (2013)(Opens in a new window), Sleeping Dogs(Opens in a new window), BioShock: Infinite(Opens in a new window), and Hitman: Absolution(Opens in a new window). All are tested at resolutions of 1080p, 1440p, and 2160p.

Top graphical presets are used in all. Tomb Raider is tested at its Ultimate setting, while Sleeping Dogs is tested at Extreme. In BioShock Infinite, we use the Ultra graphics setting with DDOF turned on. The final game test, Hitman: Absolution, also uses the top preset.

Testing Ray Tracing, DLSS, and CAS

We’ve mentioned ray-tracing, DLSS, and CAS above several times already, so a bit more on these technologies.

To test the capabilities of ray tracing and DLSS («deep learning super sampling,» a sort of AI-assisted method of anti-aliasing) in concert, we use 3DMark’s Port Royal subtest, as well as running the game Shadow of the Tomb Raider with RTX and DLSS turned on, Metro: Exodus with RTX and DLSS turned on, and Death Stranding with just DLSS turned on.

DLSS 1.0 and 2.0 are represented in these tests, with Death Stranding representing the latter. It’s a good sign for what we can expect out of DLSS moving forward, and hopefully the huge gains that cards see with the feature turned on will be expanded, with the tech’s next iterative release, to a wider number of games than the current stable of seven.

FidelityFX with CAS integration, meanwhile, is the closest thing to DLSS that either GeForce GTX or AMD Radeon card owners will get for the time being. It uses CAS, known otherwise as «contrast adaptive sharpening,» to attempt similar results as DLSS. First, it downrenders the resolution, and then using a post-processing filter, upscales the image back to something that resembles (or is hopefully indistinguishable from) the native resolution, all while adding up to 30 percent in performance in optimal scenarios.

The 3DMark Port Royal subtest works as a solid flat-out measure of how well a card’s Tensor and RTX cores are working in tandem (it returns a proprietary score, like the other 3DMark tests), while Death Stranding is more of a demonstration of how DLSS or CAS upscaling in particular can add more performance to aging cards. Shadow of the Tomb Raider and Metro: Exodus, meanwhile, are representative of how the older versions of both ray-tracing and DLSS will hold up across aging and current GeForce RTX cards.

Overclocking and Thermals

Finally, once all the standard tests have been run and we’ve gotten the numbers we need, we void the warranty and start overclocking.

PC Labs’ preferred overclocking software these days is EVGA’s Precision X1(Opens in a new window) utility, and the procedure is about as standard as it comes: Max out what we think is possible on the power limits, voltages, GPU clocks, and memory clocks, and then steadily step things down from there until the card hits a point of 100 percent stability across at least three full runs of Shadow of the Tomb Raider’s benchmark sequence.

From there, we reset the card to factory settings and turn on the FurMark stress test, using GPU-Z to record the temperature of the card for a 10-minute run, while at the same time using a FLIR One Pro thermal camera to track the waste-heat signature of the card and to see where it’s dumping the most hot air in our testbed’s case.

Get Our Best Stories!

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.

Thanks for signing up!

Your subscription has been confirmed. Keep an eye on your inbox!

Graphics Card Reviews and Lab Tests

PCMag editors select and review products independently. If you buy through affiliate links, we may earn commissions, which help support our testing.

Best Overall Graphics Cards
Best Graphics Cards for 4K Gaming
Best Graphics Cards for 1080p Gaming
Best Graphics Cards for VR
Best Cards for Compact PCs

Today’s Best Graphics Cards, Tested
By John Burek
Top Video Cards for 4K Gaming
By John Burek
How to Upgrade Your Graphics Card
By William Van Winkle
How to Run Old Games on Your Modern PC
By Chris Stobing
Top Video Cards for 1080p Play
By John Burek
How We Test Graphics Cards
By Chris Stobing

Latest Graphics Card Reviews

Asrock Intel Arc A380 Challenger ITX 6GB OC Review

Nvidia GeForce RTX 3090 Founders Edition Review

MSI Radeon RX 6750 XT Gaming X Trio 12G Review

EVGA GeForce RTX 3050 XC Black Gaming 8G Review

4.0

Editors’ Choice

Gigabyte Radeon RX 6500 XT Eagle 4G Review

XFX Speedster SWFT 210 Radeon RX 6600 Review

Gigabyte Radeon RX 6600 XT Gaming OC Pro 8G Review

Asus ROG Strix LC AMD Radeon RX 6800 XT OC Edition Review

Zotac GeForce RTX 3080 Ti AMP HoloBlack Review

Nvidia GeForce RTX 3070 Ti Founders Edition Review

4. 0

Editors’ Choice

Latest Graphics Card News

Hands On With Intel’s Arc A770 Limited Edition: A New Card Player Enters the Arena
By Michael Justin Allen Sexton

Intel’s Arc A750 Graphics Card Can Be Yours for Just $289
By Michael Kan

To Stand Out, Intel Arc GPUs Will Focus on Affordability
By Michael Kan

Intel Arc A770 Graphics Card Lands on Oct. 12 With a Surprisingly Low Price Tag
By Michael Kan

AMD Calls Out Reduced Retail Prices for Radeon Graphics Cards
By Michael Kan

Nvidia’s 4x Performance Gains for RTX 4000 GPUs Deflate Without DLSS
By Michael Kan

Nvidia CEO: High Prices for GPUs Are Here to Stay
By Michael Kan

Why No GeForce RTX 4070? It’s Coming, Nvidia Says
By Michael Kan

With Nvidia’s GeForce RTX 40 Series, Are High GPU Prices the New Normal?
By Michael Justin Allen Sexton

Portal With RTX Adds New Graphics, Full Ray-Tracing Support to the Classic Game
By Matthew Humphries

Latest Graphics Card How-Tos

How to Run Games at 4K on a 1080p Monitor With Supersampling
By Whitson Gordon

Look Out, DLSS! First Tests: AMD’s Game-Boosting Radeon Super Resolution
By Chris Stobing

What Graphics Card Do I Have?
By Whitson Gordon

New Graphics Card Too Expensive? How to Squeeze More Performance From Your GPU
By Chris Stobing

Sharpen & Speed Up Your PC Games: Testing AMD’s FSR, Nvidia’s DLSS 2. 2 & More
By Chris Stobing

What Is Ray Tracing? (And What It Means for PC Gaming)
By Charles Jefferies

Popular Graphics Card Brands

Nvidia

Latest Graphics Card Stories

A momentous product: The Arc A770 is Intel’s first flagship graphics card in more than two decades. Let’s take a look at it!

10/01/2022
By
Michael Justin Allen Sexton

Compared to Nvidia’s RTX 3060 card, Arc A750 buyers can potentially save as much as $100 or more, which is precisely how Intel is trying to stand out with its GPU line.

9/29/2022
By
Michael Kan

Intel intends to compete on the lower end of the GPU spectrum and is already busy developing its next-generation ‘Battlemage’ PC graphics cards.

9/28/2022
By
Michael Kan

Intel wants to counter the high costs of rival PC GPUs on the market, says CEO Pat Gelsinger.

9/27/2022
By
Michael Kan

The decreased prices occur as demand for GPUs has declined and AMD is preparing to launch next-gen Radeon RX 7000 products in November.

9/23/2022
By
Michael Kan

A benchmark from the company shows that performance gains become more muted if the PC game doesn’t support Nvidia’s DLSS technology.

9/22/2022
By
Michael Kan

‘The idea that a chip is going to go down in cost over time, unfortunately, is a story of the past,’ Nvidia CEO Jensen Huang says one day after the company revealed the $1,599 GeForce RTX 4090.

9/21/2022
By
Michael Kan

The company plans on going down the RTX 4000 stack over time, according to Nvidia’s CEO.

9/21/2022
By
Michael Kan

The GPU shortage is over, hurrah! But prepare to pay even more for your next-gen graphics card…yikes!

9/20/2022
By
Michael Justin Allen Sexton

The update for Valve’s 15-year-old game takes full advantage of Nvidia’s RTX 40 Series GPUs.

9/20/2022
By
Matthew Humphries

Nvidia says the RTX 4090 and 4080 deliver a two- to four-times performance increase over the RTX 3090 Ti and 3080 Ti, respectively.

9/20/2022
By
Michael Kan

The company announces the event hours before Nvidia is expected to talk up new RTX 4000 GPUs at its GTC developer event keynote.

9/20/2022
By
Michael Kan

Nvidia reportedly refused to reveal basic pre-launch information about its PC graphics card to partners like EVGA, making it diffcult for them to prepare.

9/16/2022
By
Michael Kan

Crypto-miners are shutting off their rigs and mulling selling their GPUs since few, if any, cryptocurrencies are currently profitable if you try to mine them, following the Ethereum merge.

9/16/2022
By
Michael Kan

However, miners are already eyeing two other cryptocurrencies as an alternative.

9/15/2022
By
Michael Kan

The images seem to come from a manufacturing facility for the vendor Zotac.

9/13/2022
By
Michael Kan

The most powerful model, the Arc A770, will have two variants with 8GB or 16GB of video memory.

9/08/2022
By
Michael Kan

Select graphics cards, desktops, and laptops will include a digital copy of the game until Oct. 12.

9/08/2022
By
Matthew Humphries

Meanwhile, retailers are slashing RTX 3090 Ti prices by half as they try to clear GPU inventory.

9/06/2022
By
Michael Kan

Intel’s entry into the consumer video card market, the Arc A380, is a welcome challenge to the AMD and Nvidia duopoly. But the poky performance and instability issues of this Asrock sample make it clear that Intel’s got lots of work to do to compete.

$139.99

at Newegg

(Opens in a new window)

9/02/2022
By
Michael Justin Allen Sexton

The Best Graphics Cards for VR in 2021

Decades after the 1992 flick The Lawnmower Man got our hopes up (and the ill-fated Nintendo Virtual Boy dashed them a few years later), virtual reality (VR) is finally happening in a big way. (Really, this time! We mean it.)

The simplistic, though still impressive, Google Cardboard and Samsung Gear VR platforms have both been around since 2014. And Google’s updated Daydream View landed in late 2016, upping the sophistication and comfort factor, while delivering a more robust environment for developers. These modest VR setups had you strap your smartphone into a viewer fitted with lenses, allowing your phone to become both the screen and the graphics-rendering device charged with creating an immersive world.

More About Our Picks

Nvidia GeForce RTX 3070 Founders Edition

5.0 Exemplary

Best Overall for High-Refresh AAA Gaming Plus VR

Bottom Line:

A fierce follow-up to its killer GeForce RTX 3080, Nvidia’s beautifully engineered RTX 3070 Founders Edition is a practically perfect graphics card. Gamers aiming for high-resolution, high-refresh 1440p or 4K play will find today’s best price-for-performance engine right here.

Pros

Incredibly fast for the price
Beautiful design
Great 4K gaming results
Innovative cooling system
Doesn’t run too hot

Cons

Expect a run on this card to rival the one on the RTX 3080

Read Our Nvidia GeForce RTX 3070 Founders Edition Review

AMD Radeon RX 5700 XT

4.5 Outstanding

Best for Midrange VR, Plus Play on a 1440p FreeSync Display

Bottom Line:

With its Radeon RX 5700 XT, AMD introduces its new graphics architecture, a suite of software improvements, and enough speed for the money to keep 1440p gamers very happy—and competing GeForce cards at bay.

Pros

Strong value proposition for a midrange video card
1440p results above 60fps across all benchmarks
Radeon Image Sharpening improves game visual fidelity

Cons

Only modest early returns on overclocking our sample
No hardware ray-tracing support
Loud blower cooler.

Read Our AMD Radeon RX 5700 XT Review

Nvidia GeForce RTX 3060 Ti Founders Edition

4.5 Outstanding

Good for High-Detail VR Gaming and 1440p Play on a «Budget»

Bottom Line:

If you want the best marriage of price, performance, and features for 4K and 1440p gaming, Nvidia’s GeForce RTX 3060 Ti Founders Edition is matched only by its own step-up RTX 3070 sibling. This card owns the $399 price point and delivers great value.

Pros

Beats the RTX 2080 Super in most benchmarks
Great price-to-performance ratio
Stable launch drivers
Runs cool
Short PCB, redesigned cooling system make for a compact card

Cons

RTX 3070 gives an extra margin for 4K today and tomorrow

Read Our Nvidia GeForce RTX 3060 Ti Founders Edition Review

Nvidia GeForce RTX 3080 Founders Edition

4. 5 Outstanding

Best for Valve Index Owners Who Game at High Detail

Bottom Line:

If you’re a current (or aspiring) 4K gamer, Nvidia’s ferocious, field-redefining GeForce RTX 3080 graphics card is the only one worth considering.

Pros

Chart-topping performance for the price
Whisper-quiet operation
Innovative cooling system
Gorgeous shroud design
Tons of complimentary software features

Cons

No DLSS 3.0 (yet)
Not much overclocking headroom
Gains in 1080p and 1440p are less impressive than in 4K

Read Our Nvidia GeForce RTX 3080 Founders Edition Review

AMD Radeon RX 5700

4.0 Excellent

A Cheaper Alternative to the Radeon RX 5700 XT

Bottom Line:

Gaming at 1440p is the strength of the Radeon RX 5700, a solid entry that, alongside its RX 5700 XT kin, makes AMD a real player again the field of midrange video cards.

Pros

Excellent price-to-performance ratio.
Solid results in 1440p gaming.
Blower cooler vents most heat out your PC case.

Cons

Blower cooling is still pretty loud.
Overclocking ceiling is low, at this writing.

Read Our AMD Radeon RX 5700 Review

MSI GeForce GTX 1660 Ti Gaming X 6G

4.0 Excellent

Ground-Floor VR With a Current-Gen Card

Bottom Line:

MSI’s GeForce GTX 1660 Ti Gaming X 6G is exceptional at doing what it’s designed to do: deliver a moderate-cost GPU option to gamers in search of high refresh rates in 1080p.

Pros

Great price-to-performance ratio for 1080p gaming.
Beats previous-generation GTX cards on both sides of its price.
Exceptional cooling.
Solid overclocking potential.

Cons

Not great for 4K gaming.
MSI’s upclocked card closer in pricing to RTX 2060 Founders Edition than most GTX 1660 Ti cards.

Read Our MSI GeForce GTX 1660 Ti Gaming X 6G Review

Nvidia GeForce RTX 2060 Super

4.0 Excellent

Best for VR Plus High-Refresh-Rate Esports

Bottom Line:

Nvidia’s GeForce RTX 2060 Super straddles the fine line between a video card made for esports and one for AAA gaming performance, and delivers consistently on each side where it matters most.

Pros

Strong results in high-refresh gaming and 1440p.
Competitively priced.
Highly attractive card build.
DVI port for legacy monitors.

Cons

Not powerful enough to handle all AAA games at higher resolutions.

Read Our Nvidia GeForce RTX 2060 Super Review

AMD Radeon RX 6800 XT

3.5 Good

Best for High-Detail Gaming, Plus VR, for the AMD Faithful

Bottom Line:

AMD’s Radeon RX 6800 XT impresses in spots with record-breaking results in specific benchmarks, but inconsistent frame rates on several games, and driver-stability issues, keep it from toppling the best of Nvidia’s GeForce RTX.

Pros

Lots of overclocking headroom
Laden with new features
Beats competing Nvidia GeForce RTX options in select benchmarks

Cons

Inconsistent results across multiple games
Driver-stability issues in the early going
Runs hotter than most GPUs

Read Our AMD Radeon RX 6800 XT Review

AMD Radeon RX 6700 XT

3.5 Good

Best for Gamers Who Stick to Recent Titles

Bottom Line:

AMD’s Radeon RX 6700 XT is a solid competitor to Nvidia’s lineup of midrange GPUs—if you stick to recent, optimized titles.

Pros

Competitive frame rates, in games that are optimized
SAM testing sees frame-rate gains in select titles
Debut MSRP is $20 below that of Nvidia’s GeForce RTX 3070

Cons

DirectX 11-based games may run significantly slower than on competing Nvidia cards
Higher temperatures under stress than competing Nvidia cards
Radeon reference design is underwhelming compared with Nvidia’s RTX 30 Series Founders Editions
Coil whine under load in our sample card

Read Our AMD Radeon RX 6700 XT Review

EVGA GeForce RTX 3060 XC Black Gaming 12G

3. 5 Good

Good for VR Players Building Out a Compact PC

Bottom Line:

Suited to 1080p and 1440p gaming, EVGA’s XC Black Gaming 12G version of the GeForce RTX 3060 is an able performer, though Nvidia’s own RTX 3060 Ti outshines it on value.

Pros

Fast frame rates in multiplayer titles
Compact card design
Reasonable overclocking performance

Cons

Nontrivial performance drop versus the GeForce RTX 3060 Ti
Minor gains over previous (RTX 2060) generation
Pricey for its performance class

Read Our EVGA GeForce RTX 3060 XC Black Gaming 12G Review

These were impressive early attempts at «cheap VR,» but the screens on most smartphones just don’t pack enough pixels to deliver a sharp image a few inches from your eyeballs. And even since then, today’s smartphone graphics chips aren’t quite up to the task of rendering complex 3D worlds with high-resolution textures in a way that passes muster up that close. It’s tough to feel truly immersed in a virtual world when you’re staring at a grainy screen, blocky text, and distant mountains that look pixelated.

Those who want their virtual reality to look a little more, well, «real» will be more interested in today’s powerful mainstream VR headsets, or head-mounted displays (HMDs)—the Oculus Rift (and the Oculus Rift S), the HTC Vive, the HTC Vive Pro, and the Valve Index—as well as the ‘tweener category of Windows Mixed Reality headsets. The Oculus and HTC HMDs require a fairly powerful PC to create lush environments right in front of your eyes.

Here’s how to assess what you need to use them. Spoiler alert, though, in case the headline wasn’t a big enough clue: The graphics card matters. A lot.

First: A Look at the Headset Specs

The Vive and Rift have rather similar core specs, and as a result, have similar minimum hardware demands.

Both have an effective resolution of 2,160 by 1,200 pixels (that is, 1,080 by 1,200 pixels per eye) and a refresh rate of 90Hz. And both use OLED screens for rich blacks and vivid colors, just like Samsung’s high-end phones and tablets.

The Valve Index, meanwhile, has an even higher resolution. Dual 3.5-inch 1,440-by-1,600-pixel AMOLED screens drive its display system, for an effective resolution of 2,880 by 1,600 pixels. The result? A significantly sharper picture, with a remarkably high pixel density of 615 pixels per inch. At $500 just for the headset ($1,000 for the full kit), it’s awfully expensive, though. Then there’s the Vive Cosmos family, with slightly higher-res 1,440-by-1,700-pixel screens.

Beyond the HTC, Valve, and Oculus offerings are a handful of so-called Windows Mixed Reality headsets, which combine aspects of VR with reality-overlay augmented reality (AR). They haven’t really taken off on the consumer side, but this category of devices includes the Samsung HMD Odyssey, which has the same resolution and same refresh rate as the Vive Pro, and the Dell Visor, with dual 1,440-by-1,440-pixel screens and a 90Hz refresh rate. They have lower hardware requirements than the Valve, HTC, and Oculus headsets, but these requirements can vary according to their intended usage cases. They’re not entirely about immersive gaming (not all VR games work on these headsets) and can be used in situations involving interactive training or other specialized work usage cases.

What Your VR PC Needs: The Minimums

If you’re putting together a gaming PC now to use with one of these headsets, or updating your existing desktop to make sure it’s ready for VR, just what will you need? The short answer: a fair bit of graphics muscle.

The resolutions of the Rift and non-Pro Vive headsets aren’t exactly ground-breaking, especially for PC gamers who are already gaming at the much higher resolutions of 1440p (2,560 by 1,440 pixels) or 4K (3,840 by 2,160 pixels). But the 90Hz refresh rate that VR headsets demand means that your games will have to be running at or close to 90 frames per second (fps), or ideally higher, to look as smooth as they should. Don’t downplay the necessity of that: Frame-rate smoothness matters a lot more in VR than it does in traditional gaming, because judder and screen tearing while you’re moving your head around in a virtual world can cause dizziness and nausea.

The minimum hardware requirements for the HTC Vive and Vive Pro include an Intel Core i5-4590 or an AMD FX-8350 or better processor. In general, that means any recent-model Core i5 or Core i7 desktop CPU, such as the Intel Core i5-11600K or one of AMD’s Ryzen 7 chips, should be more than sufficient. Given the similarity of the screens in the original Vive and the Rift, it’s no surprise that Oculus’ published minimum recommendations for the Rift include the same Intel CPU suggestions. For either HMD, you’ll ideally also want 16GB or more of RAM, though 4GB is given as the bare minimum for the Vive and 8GB for the Rift. If you don’t want to be standing for long stretches in virtual darkness, waiting for your game’s levels to load, we’d also strongly recommend opting for a solid-state drive (SSD) in any VR-ready system.

The graphics card is the key piece of hardware, though. If you have an older video card, for use with the Index, original Vive, or Rift (and the Rift S), you’ll really want at least a GeForce GTX 1060 card from Nvidia, or a Radeon RX 580, Radeon RX 590, or Radeon RX Vega card from AMD. The GeForce GTX 1050 and GTX 1050 Ti won’t cut it for the Vive, though an even older-gen GeForce GTX 970/980 or Radeon R9 290/390 might just squeak you by. The GeForce GTX 1050 Ti is given as the bare minimum for the Rift and the Rift S. With newer cards, a Radeon RX 5600 Series or a GeForce GTX 1660 Ti is where you want to start. Or any of the GeForce RTX cards will do. (More on those in a moment.)

For the Vive Pro or newer Vive Cosmos HMDs, you’ll want to look a little bit higher up the card stacks; the squeak-by cards are the same as for the regular Vive, but the recommended baseline is higher. More on that in a moment.

Up the Stack: Higher-End GPUs for VR

If you’re strapped for cash, the abovementioned «borderline» cards should deliver playable performance for all but the most demanding current flagship VR titles.

But given that these cards are the baseline recommended cards, it’s likely, down the road, that a PC built around one of them might not run all future VR games perfectly at their highest settings. And if you’re building or upgrading a PC specifically for VR, you want to be sure that you’re allowing for some performance overhead, so your system can play VR titles that haven’t hit the market yet, not to mention hot AAA games outside of VR at your resolution of choice.

For that extra breathing room, you’ll want a more powerful graphics card than the threshold GeForce GTX 1060 or Radeon RX 480/580 cards. Traditionally, those looking for peak possible non-VR gaming performance have sometimes opted for a multi-card setup with two or more high-end GPUs. You could do that for VR, but the roster of VR titles that properly support SLI (Nvidia) or CrossFire (AMD) multi-card configurations is small, and support for SLI has almost evaporated in the last two generations of GeForce cards (the GeForce RTX 20 and 30 series). This, combined with the potential for frame-timing wobbles and other issues that plague multi-card setups in in general, makes a single powerful graphics card by far your best bet for smooth VR at the moment.

So, then, which powerful cards? As 2021 kicks off, Nvidia cards based around Nvidia’s «Pascal» 10-series graphics processors have yielded the floor to the company’s GeForce RTX GPUs, in both the «Turing»-based RTX 20 Series as well as the newer «Ampere»-based RTX 30 Series graphics cards.

The VR-friendly GeForce RTX line is currently spearheaded by the flagship GeForce RTX 3080 and runs down, at the moment, to the more budget-minded GeForce RTX 3060 Ti and RTX 3060. Any of the GeForce RTX cards released to date are more than good enough for VR gaming in the state it’s in today.

So are AMD’s top 5000 series offerings, the Radeon RX 5700 and RX 5700 XT, as well as the newest generation of RDNA 2 cards, the Radeon RX 6800 and RX 6800 XT. The AMD Radeon RX 6700 XT is also a strong contender for AMD, though its somewhat lower frame rates in certain games that use the DirectX11 API might give some legacy VR players some pause.

If these top-enders are out of your price range, but you’re still looking for something that’s a step above the VR GPU baseline, a good middle-ground card is one based on the GeForce RTX 3060, or even the GeForce RTX 2060, if you can still find one on the market. (All of the 10 Series cards higher than the GeForce GTX 1060 are end of life, in favor of the high-end GeForce RTX cards and the GeForce GTX 1660 and GTX 1660 Ti, which are the new baseline for VR among current Nvidia cards.)

Note: The GTX 1070 is also HTC’s recommended baseline card for use with the Vive Pro HMD, so if you’re looking for a card to use with that headset, you ought to be in this premium-card mindset to start with.

So, Which Card to Buy?

To be sure: Here in 2021, video card prices are high, and availability is spotty as the industry works through its shortages of GPU production and supply bandwidth. But know that the more overlooked mainstream cards in AMD’s and Nvidia’s lines (the higher-end GeForce GTX, as opposed to RTX cards, and the RX 5700 Series Radeons) can get you in on the VR action without spending four figures from an eBay scalper. You don’t necessarily need the very latest and greatest card, unless you also aim to play AAA games outside your VR headset at resolutions like 1440p or 4K.

One last item to think about in light of VR on the latest video cards, before we get to our card picks: Many cards based on the GeForce RTX 2080 Ti, RTX 2080, and RTX 2070 GPUs, and some RTX 2060 cards (including the «RTX Super» equivalents for models that have them), have on their backplane a port called VirtualLink. (That’s not to be confused with NVLink, the edge connector now used for bridging a few choice RTX GeForce cards for twin-card SLI.) VirtualLink ports are actually special USB Type-C connectors that also support DisplayPort output and a certain amount of power delivery; the VirtualLink spec outlines an operating mode that allows for data, power, and video to future VR HMDs to run across this single-wire connection.

That said, that future may never arrive. No current VR HMDs support VirtualLink at this writing, nor are we aware of any coming, but it’s something to keep in mind if you’ve seen mention of it. (Valve planned a VirtualLink adapter for the Index and later canned it.) Both Nvidia’s latest RTX 30 Series line and AMD’s Radeon RX 6000 series have done away with the VirtualLink port altogether on the reference cards, which may be cementing VirtualLink as yet another «come and gone» technology before it even had a chance to get off the ground. There’s been no VirtualLink action in new 2021 cards, either.

As for the cards below, bear in mind that the AMD Radeon RX cards and Nvidia GeForce Founders Edition cards we tested in 2020 are reference cards, or the chip makers’ baseline versions of cards with their GPUs. Many other card makers issue cards, not reviewed here, with the same core graphics processor that will deliver close or better performance. You can consider those equivalents in the same class, as well.

The Best Graphics Cards for 4K Gaming in 2022

Editor’s Note: Before you dive into this guide, as you’ll see from many of the prices above, the availability and pricing situation for GPUs is anything but «normal» right now, and has been skewed since early on in the pandemic. If you plan to buy a card soon, also see this buying-strategies guide for advice on finding cards at a fair price. If you want to wait it out a bit longer, check out this how-to tutorial on getting the most performance from the GPU you already own.

The Best Graphics Card Deals This Week*

*Deals are selected by our commerce team

MSI Geforce RTX 3090 24GB Graphics Card
(Opens in a new window)
— $1,762.99
MSI Geforce RTX 3090 Ti 24GB Graphics Card
(Opens in a new window)
— $1,269.60

(List Price $1,400)
EVGA GeForce RTX 3090 FTW3 Ultra Graphics Card
(Opens in a new window)
— $1,126.99

(List Price $1,919.99)
Zotac Gaming RTX 3060 Twin Edge OC 12GB Graphics Card
(Opens in a new window)
— $420.57

(List Price $549.99)
Zotac Gaming RTX 3080 Ti Trinity OC Graphic Card
(Opens in a new window)
— $869.99

(List Price $1,149. 99)

More About Our Picks

Nvidia GeForce RTX 3080 Founders Edition

4.5 Outstanding

Best Overall Graphics Card for Top-Detail 4K Play

Bottom Line:

If you’re a current (or aspiring) 4K gamer, Nvidia’s ferocious, field-redefining GeForce RTX 3080 graphics card is the only one worth considering.

Pros

Chart-topping performance for the price
Whisper-quiet operation
Innovative cooling system
Gorgeous shroud design
Tons of complimentary software features

Cons

No DLSS 3.0 (yet)
Not much overclocking headroom
Gains in 1080p and 1440p are less impressive than in 4K

Read Our Nvidia GeForce RTX 3080 Founders Edition Review

Nvidia GeForce RTX 2080 Ti Founders Edition

4.0 Excellent

A Solid Last-Gen Alternative to GeForce RTX 3080

Bottom Line:

A Ferrari among gaming GPUs, Nvidia’s GeForce RTX 2080 Ti Founders Edition represents the fastest class of cards that money can buy. Just be ready to pay a supercar price to enjoy its luxury ride for 4K and high-refresh gaming.

Pros

Sets a new bar for single-GPU performance.
Quiet, cool-running design.
Supports ray tracing and DLSS for future games.
Easy to attain at least modest overclocks.

Cons

Founders Edition commands a $200 premium over an already expensive base/reference card.
Games will take time to adopt ray tracing and DLSS.

Read Our Nvidia GeForce RTX 2080 Ti Founders Edition Review

Nvidia GeForce RTX 3070 Founders Edition

5.0 Exemplary

Best «Value» Graphics Card for 4K Play

Bottom Line:

Pros

Incredibly fast for the price
Beautiful design
Great 4K gaming results
Innovative cooling system
Doesn’t run too hot

Cons

Expect a run on this card to rival the one on the RTX 3080

Read Our Nvidia GeForce RTX 3070 Founders Edition Review

Nvidia GeForce RTX 3060 Ti Founders Edition

4.5 Outstanding

A Solid Alternative to GeForce RTX 3070

Bottom Line:

Pros

Beats the RTX 2080 Super in most benchmarks
Great price-to-performance ratio
Stable launch drivers
Runs cool
Short PCB, redesigned cooling system make for a compact card

Cons

RTX 3070 gives an extra margin for 4K today and tomorrow

Read Our Nvidia GeForce RTX 3060 Ti Founders Edition Review

Nvidia GeForce RTX 2070 Super

4. 5 Outstanding

Best Graphics Card for 4K Play on 60Hz Displays

Bottom Line:

Packing near-RTX 2080 performance and similar specs, Nvidia’s GeForce RTX 2070 Super rules for solid 4K play at 60Hz and high-refresh gameplay at 1440p. It’s a killer card for the money.

Pros

Superb price for performance.
Stable overclocking results help it rival original RTX 2080.
Runs quiet.
RT and Tensor cores ensure future-proofing.

Cons

It didn’t come sooner.

Read Our Nvidia GeForce RTX 2070 Super Review

MSI GeForce RTX 3090 Suprim X 24G

4.0 Excellent

Best Graphics Card for 4K Play, Cost No Object

Bottom Line:

The MSI GeForce RTX 3090 Suprim X 24G is (very) large and (very) in charge, returning the fastest results we’ve ever seen out of a GPU, albeit for an eye-bugging price.

Pros

Hands down the fastest consumer GPU we’ve ever tested
Quiet running
Polished, refined design

Cons

Very expensive, even versus Nvidia’s base-model RTX 3090 Founders Edition
Power-hungry
Requires three power connectors
Three-slot width, 13-inch length
Minimal overclocking headroom

Read Our MSI GeForce RTX 3090 Suprim X 24G Review

AMD Radeon RX 6800 XT

3. 5 Good

Best Top-End AMD Graphics Card for 4K Play

Bottom Line:

Pros

Lots of overclocking headroom
Laden with new features
Beats competing Nvidia GeForce RTX options in select benchmarks

Cons

Inconsistent results across multiple games
Driver-stability issues in the early going
Runs hotter than most GPUs

Read Our AMD Radeon RX 6800 XT Review

AMD Radeon RX 6800

3.5 Good

Best AMD Graphics Card for 4K Play at Medium Detail

Bottom Line:

While AMD’s Radeon RX 6800 posts some impressive results in certain benchmarks, its uneven performance in others makes it a difficult Day One graphics-card recommendation versus Nvidia’s GeForce RTX 3070.

Pros

High overclocking ceiling
A bevy of new features to play with
Cooler temperatures than Radeon RX 6800 XT

Cons

Subpar results with AAA legacy games
Expensive versus Nvidia competition
Limited ray-tracing support

Read Our AMD Radeon RX 6800 Review

AMD Radeon RX 6700 XT

3.5 Good

A Solid Alternative to Radeon RX 6800

Bottom Line:

AMD’s Radeon RX 6700 XT is a solid competitor to Nvidia’s lineup of midrange GPUs—if you stick to recent, optimized titles.

Pros

Competitive frame rates, in games that are optimized
SAM testing sees frame-rate gains in select titles
Debut MSRP is $20 below that of Nvidia’s GeForce RTX 3070

Cons

DirectX 11-based games may run significantly slower than on competing Nvidia cards
Higher temperatures under stress than competing Nvidia cards
Radeon reference design is underwhelming compared with Nvidia’s RTX 30 Series Founders Editions
Coil whine under load in our sample card

Read Our AMD Radeon RX 6700 XT Review

Thanks to streaming services such as Netflix, Amazon Prime, and even YouTube, it’s finally getting easier to find actual 4K (also referred to as «Ultra HD») video content. But as awesome as 4K video looks, if you’re aiming to immerse yourself in a pixel-dense world, it’s hard to beat playing cutting-edge PC games in 4K. And from a hardware perspective, that’s a much more daunting prospect.

Only the latest consoles—the Sony PlayStation 4 Pro, the Microsoft Xbox One X, the PlayStation 5, and the Xbox One Series X—will output games at 4K. But really, if you want to play brand-new AAA games at 4K with the best visuals, you’ll need a desktop PC equipped with a very powerful graphics card—especially if you want your in-game eye candy dialed all the way up. After all, if you’re investing in a 4K monitor or a 4K TV for gaming, you want things to look as good as they can. Running games at 4K resolution but dialing down the detail and effects settings in your games is working at cross-purposes. So the PC graphics card you buy matters—a lot.

4K Gaming: High-End Cards and Dual-GPU

At the moment, to deliver smooth frame rates at high settings at 4K resolution on a PC (that’s 3,840 by 2,160 pixels, for the record) with the most-demanding games, you’ll need to opt for one of the most powerful consumer-grade graphics cards available. These days, those cards include Nvidia’s «Ampere»-architecture GeForce RTX 3080 Founders Edition and the one-step-up GeForce RTX 3090 Founders Edition, (right in the middle we have the RTX 3080 Ti Founders Edition), or one of the many custom-cooled and/or overclocked models based on these cards’ GeForce RTX 3080 or RTX 3090 graphics processors (GPUs). The GeForce RTX 3070 is also an excellent choice for one-step down play from a RTX 3080 that’s still capable at 4K, while the more recently released seat filler between those two, the GeForce RTX 3070 Ti, is a bit expensive for its capabilities at $599 list price but hits 60fps at 4K more reliably than its predecessor.

Also in play for 4K: Cards based on AMD’s mid-2019 addition to its lineup, the Radeon RX 5700 XT (as long as you don’t mind turning down your settings a bit), as well as both as both cards featured in the new Radeon RX 6000 Series, the Radeon RX 6800 XT and the RX 6800, as well as the AMD Radeon RX 6700 XT. But none of these cards, GeForce or Radeon, comes cheap, most of them starting around $400 list price for the base models of the RX 5700 XT, and as high as $1,799 (!) for some third-party overclocked RTX 3090 cards, not even counting what price gougers add to the list prices these days.

(Photo: Zlata Ivleva)

You could also, in theory, pick up two GeForce RTX 3090 cards and use them in a paired NVLink arrangement to push well above 60fps.On the opposite tack, you can scrape by with a single GeForce RTX 3070, or a GTX 3060 Ti for around $400, again if you are lucky enough to find one at list price.

Note that NVLink is only possible now with the very expensive GeForce RTX 3090. In some games, an NVLink setup should deliver better gaming performance than a single RTX 3080 or RTX 3090 card. Note, though, that if you do go this two-card route, multi-graphics setups can introduce side issues. Most games don’t ship on launch day with the optimizations to take advantage of multiple-card graphics, and some games never deliver multi-graphics support at all. We also expect multi-card support to fade faster in 2022 and beyond, as support for it is limited to such elite cards.

(Photo: Zlata Ivleva)

For this reason, if you’re the kind of enthusiast PC gamer who likes to jump on games on the day they’re released, multi-GPU setups aren’t ideal solutions. Also, you might run across issues with frame timing, in which onscreen game frames don’t get delivered exactly in sync, resulting in a subpar experience. For this reason, we recommend buying the best single card for the performance level you’re after, whenever possible.

If money is no object, that single card is the $1,499-MSRP Nvidia GeForce RTX 3090, which might as well be in a class all its own (though it more appropriately belongs alongside content creator-focused cards like the Titan RTX). The RTX 3090 has been found to be, on average, only around 10% faster in gaming than the $699 GeForce RTX 3080, despite costing nearly twice as much for the Founders Edition card. Maybe it’s for this reason that the RTX 3090 has been billed by Nvidia as «the world’s-first 8K gaming GPU,» and early tests show while that claim isn’t a stretch, 8K gaming is so far beyond feasible for 99.9% of buyers right now that it’s effectively a moot point.

These cards are made for much more than gaming, deployed more often in creative fields that do a lot of 4K and 8K video editing, 3D rendering, or 3D modeling. In a price-for-performance sense, they’re major overkill for games, and they are often not optimized to take advantage of top titles as efficiently as the gaming-centric GeForce RTX 3080 cards (and its lessers) are.

So yes, while technically the GeForce RTX 3090 could push 4K-gaming frames with grace, at that price you’re better off going with a single RTX 3080 and spending the leftover $700 or $800 upgrading your RAM, CPU, and motherboard at the same time, or saving the difference. Again, this assumes you can find one of these cards at close to list price.

4K Gaming Cards: The Best «Budget» Options

If your budget can’t quite bear laying out $500 or more for a graphics card (though you don’t exactly have much of a choice otherwise, these days), you can find some less-expensive cards that can handle 4K gaming at lower settings. You won’t get the absolute best visuals possible, but 4K gaming is technically attainable.

If you don’t mind running games closer to medium detail settings at 4K, but you still want to experience the pixel-dense glory of games running at 3,840 by 2,160, the last-generation AMD Radeon RX 5700 XT and the GeForce RTX 2060 Super are both still capable engines, and actually have a higher chance of being in stock than more recently released options from either company. Just remember that you won’t be able to play many games at the highest detail settings unless settings like DLSS or an image sharpener are activated.

(Photo: Zlata Ivleva)

The card set to replace both the RTX 2070 and the RTX 2070 Super, the Editors’ Choice-winning Nvidia GeForce RTX 3070 Founders Edition ($499 MSRP), blows out the value proposition of what smooth 4K gaming has cost up until now. In our tests of that card, we found it regularly pushed even the most demanding titles just up to, or well over, the 60fps/4K mark in the highest detail settings. It sits in our eyes as the first true «value» pick for 4K gamers. If you’re only trying to game on a 4K/60Hz display (as many living-room gamers might), there isn’t any other option that performs to that spec at this price point. AMD’s Radeon RX 6800 matches or beats the GeForce RTX 3070 FE regularly enough to be included in this list, however it just grazes the line of the mainstream, and edges slightly into the high end at $579 MSRP, competing with the Nvidia GeForce RTX 3070 Ti Founders Edition at $599. Again MSRP may be your lucky day or a fantasy here in early 2022.

Overall, we can’t recommend going much lower than these cards today if you’re serious about 4K gaming, though. One of the biggest concerns that any cost-conscious PC gamer should have when choosing new hardware is how «future-proof» a card is, and given that cards below the GeForce RTX 3080 barely scratch the surface of pushing 60 frames per second (fps) on most current titles at middling settings, that viability will only continue to drop for new games released later this year.

Aside from gauging raw performance, you should keep a few other factors in mind when shopping for a powerful 4K-capable graphics card. Let’s run through these one by one.

Consider the Target Display

The first consideration? The particular 4K display you’ll be using. If you opt for a 4K monitor with a DisplayPort 1.4 input (which has the bandwidth to deliver 4K content at 144Hz, or up to 144 frames per second), any of the current-gen cards here should serve you well. But if you are thinking of using a 4K television as a large-screen gaming display, you’ll likely be stuck using HDMI 2.0 or HDMI 2.1 to jack in.

Why? Most 4K TVs lack DisplayPort connectors. This locks you into using HDMI, and thus to 4K at 60Hz. This doesn’t actually matter all that much outside of the world of the unusual Nvidia BFGD (short for «Big Format Gaming Display») monitors that will run 4K at up to 120Hz. Though it is very much game-dependent, today no single GPU will push AAA games at max settings in 4K much beyond 60 frames per second natively unless tech like DLSS or image sharpening is in use. (More on that in a bit.)

Beyond TVs, though, gaming monitors have seen more evolution in the last two years than in practically the whole decade before. High-refresh-rate (that is, above 60Hz) 4K monitors are now an option and available from multiple makers. Yes, some are more expensive than a beater of a used car (averaging anywhere between $800 to $2,500 at this writing). However, if you have that kind of cash to throw around, they are worth a look.

(Photo: Zlata Ivleva)

Options like the Asus ROG Swift PG27UQ(Opens in a new window) and the Acer Predator XB3 are typical of this market, and both provide 4K screens that can be boosted as high as 144Hz under specific conditions. We say «specific,» because as of this writing the two main cables that carry the signal to the monitor (DisplayPort 1.4b and HDMI 2.1) are only capable of delivering a full 4:4:4 signal up to 120Hz. To get to that holy grail of 144Hz, monitor manufacturers use a workaround known as «chroma subsampling,» which brings the color palette down to 4:2:2. That is fine for movies and some 3D gaming, but it can wreak havoc on content like text or graphical elements rendered in the OS.

Why do we mention all that? Because it might be tempting to break the piggy bank, rush out, grab a pair of GeForce RTX 2080 Ti or RTX 3090 cards (if you can find two at any price!), throw them into an NVLink configuration, and try to tap into 140-plus frame-per-second rates at 4K on ultra-high detail settings. Realistically, with one of these elite monitors, you should settle for a setup capable of pushing closer to 120fps (120Hz) to get the best visual results from your games and daily computer usage.

That’s pie-in-the-sky for almost all buyers, though. Bottom line: Unless you’re really awash in dough, you’ll be just fine with a monitor or a TV that pushes 4K games at 60Hz. That’s near the top frame rate that today’s rigs equipped with a single GeForce RTX 3080 card will be able to achieve on leading games, anyway. Leave higher-end frame-rate aims in 4K to Powerball winners.

One last consideration: high dynamic range (HDR). It looks gorgeous when implemented properly, and many new games (as well as a back catalog of older games) support the spec. You’ll find many 4K HDR gaming monitors already on the market to choose among, and most have come down in price far enough to keep them competitive with their non-HDR counterparts. From the point of view of card buying, though, no worries: All current-gen GeForce and Radeon cards support it.

How Much Video Memory Is Necessary?

The other thing to watch for while shopping for a 4K-ready gaming card is the amount of dedicated video memory on the card. Generally, 4GB of memory is plenty if you’re gaming at 1080p or below, but when you step up to 4K, a graphics card needs to handle much more data.

To keep your gaming sessions running smoothly at 4K and high detail settings, you’ll want a card with at least 6GB of memory. A card with 8GB of GDDR6 or more is your best bet, especially if you’re the type who likes to download game mods and/or high-resolution texture packs, which are sometimes specifically created to deliver a greater level of in-game detail for high-end cards that have extra memory capacity.

The RTX 3080, being the luxury-ride card that it is, pushes things to the limit in this category. Every card comes with a whopping 10GB of onboard GDDR6X memory, capable of pushing a staggering 616GB per second across a 352-bit width bus. And the AMD Radeon VII packs even more, 16GB of HBM2 memory, which is more of a boon for content creators than for your average gamer. (The 24GB on the RTX 3090 falls into the same category as the Radeon VII’s 16GB.)

Again, pretty much anything around the 6GB level will hold its own on medium-detail 4K games, but start diving into the high/ultra presets of cutting-edge games, and you’ll need all you can get.

DLSS, Image Sharpening, and Super Resolution: 4K Gaming’s (Potential) Ace in the Hole?

DLSS, or «deep learning super sampling,» is a technology developed by Nvidia for use in its latest lineup of RTX cards. The tech, which uses an AI architecture to streamline the process of anti-aliasing, offers significant performance boosts over non-DLSS results.

For example, at 4K resolution with DLSS turned on, an RTX 3080 Founders Edition racked up an incredible improvement of nearly 60% with the feature enabled in our Death Stranding benchmark. This made it so even low-end cards like the GeForce RTX 3060 Ti could comfortably run the game above that coveted 60fps mark. (It achieved an average of 63fps in native, while turning on DLSS in «Performance Mode» boosted that number to 105fps.) DLSS is a huge leap in gaming tech that brings 4K-playing affordability down to the midrange, and if it can be implemented on a wider basis (rumors say all games that support TAA may see DLSS integration in the near future), it will make it possible for card owners in the $399 bracket to play AAA titles at the highest detail levels, all in 4K.

DLSS indeed shows major promise in games, but we use the term «games» loosely for now, given that, as of this writing, only a limited number of them are trained in how to utilize it. A new version, DLSS 2.0, introduced in spring of 2020, launched with support for just a handful of games, along with promises of easier game-developer implementation. All this is to say that while DLSS is certainly impressive (it may even be the one thing that brings the world of 4K gaming into the hands of mainstream gamers), right now any widespread implementation still looks to be a ways off.

AMD has also posited its own resolution enhancers under the name Super Resolution, and AMD owners can still utilize what we consider to be the best of the various «image sharpening» technologies out there, Radeon Image Sharpening (RIS). As a hardware-based sharpener, RIS gives the best results when users downscale the resolution of their game, and then re-sharpen it using either RIS, or other options like Nvidia’s Freestyle or the third-party program ReShade.

When tuned just right, we’ve found that certain games can add up to a 30% improvement in frame rates without taking a hit in visual quality. This means that midrange cards like the Radeon RX 5700 XT can more reliably hit 4K/60fps, and you can save money on your hardware without sacrificing graphical fidelity.

Understanding Card Length, Power Requirements

If you’re rocking a full-tower PC, card size is probably not an issue. However most high-end, 4K-capable video cards are three slots wide (two at a minimum) and a little more or less than 10. 5 inches long, which means that most MicroATX chassis won’t fit these monsters. Even many midsize ATX cases may find them a squeeze.

(Photo: Zlata Ivleva)

That said, there are some cards that make compromises, like the MSI GeForce RTX 2060 Aero ITX OC(Opens in a new window). This petite powerhouse takes only two card slots and measures just 7 inches long. But the GeForce RTX 2060 GPU itself will only be able to push around 60fps at medium settings for 4K gaming, at best, unless DLSS is turned on.

As for the power requirements, if your existing system already has a high-end video card in place, you’ll likely be fine, but check the recommended minimum power-supply wattage for any 4K-capable card you’re considering. Generally, an 800-watt supply should keep most any current single 4K-capable card juiced; the wattage minimum can vary from card to card.

Many of the latest top-end cards also require two, sometimes even three power-supply leads (six- or eight-pin, likely some combination of both) from your supply, so make sure you have the proper cabling in place or adapters on hand. Note: The RTX 3080 and RTX 3090, in their Founders Edition guises, employ an adapter from a new 12-pin power socket to typical six- and eight-pin connectors.

Should I Get a Card That’s Overclocked Out of the Box?

The distinctions among the many high-end third-party cards that are capable of 4K play can be esoteric. One of the big ones, though, is the presence (or not) of enhanced cooling hardware on the card to handle user-initiated overclocking of the GPU, or sometimes even overclocking done at the factory. Reviews of individual cards will get down into the weeds of exact clock rates or factory overclocking. But know that an overclocking focus is often a key reason why some cards of the same class (such as different RTX 3080 cards) vary so much in price.

(Photo: Zlata Ivleva)

Overclocking-minded cards tend to be larger than their same-GPU kin, with more fans and/or more elaborate heat pipes and sinks. The most expensive cards in a given line tend to be the ones with the beefiest hardware for overclocking or with a factory overclock done out of the box. Telltale cards of this kind include Zotac’s AMP Extreme Series, MSI’s Gaming X and Gaming Z, Gigabyte’s WindForce and Xtreme, Asus’ Republic of Gamers, and EVGA’s FTW series.

Some Basic 4K-Gaming Benchmarks…

Want some real world numbers? We’ve benchmarked all the cards recommended on this list, and here’s a snapshot of how they perform in a handful of AAA titles at 4K…

As you can tell from these graphs, the GeForce RTX 2080 Super, the GeForce RTX 3060 Ti, the GeForce RTX 3070, and the Radeon RX 6700 XT make up one roughly equivalent tier, with the GeForce RTX 3080, Radeon RX 6800 XT, GeForce RTX 2080 Ti, and MSI GeForce RTX 3090 Suprim X 24G on planes of their own. Another interesting story being told here, though, is how many cards benefit from the application of either DLSS or CAS. Midrange cards like the RTX 3060 Ti are nearly able to push 4K/60fps on Shadow of the Tomb Raider at its highest detail setting when both RTX and DLSS are turned on, while games like Death Stranding show the true capabilities of CAS and DLSS when pushed to their pre-tuned limits.

(Note: Those Death Stranding results were recorded with DLSS «Performance» mode turned on, and Kojima Studios has since added «Ultra Performance» mode to the settings menu after we started testing. That’s right—this game can go even faster than the results we’ve recorded above!)

But one thing is clear: At the moment, if you can afford (and can find!) a GeForce RTX 3080 or an AMD Radeon RX 6800 XT, that’s the card you want.

So, Which Card Should I Buy for 4K Gaming?

We’ve tested samples of all the key video cards in Nvidia’s «Turing» and «Ampere» and AMD’s «Polaris,» «Navi,» and «Big Navi» families. Below are our current favorites for 4K gaming. Thanks to the tech of image sharpeners and DLSS, which have only continued to gain traction with developers this year, the options for budget buyers to enter the 4K gaming space are broader than ever. For purists, though, mega-monsters like the Nvidia GeForce RTX 3080 Founders Edition and the AMD Radeon RX 6800 XT are bringing heaps of power to the table, and pushing even the most demanding titles well past any frame rates we’ve seen before.

As it turns out, 2021 was a great year for 4K gamers (objectively speaking, and not considering the current shambolic state of the GPU marketplace), and doubly so if any of your favorite games support DLSS or CAS. As monitor makers continue to crank up the peak refresh rate of 4K displays, by the mid-2020s we could be wondering how we ever «only» played at 4K/60fps. But until then, both Nvidia and AMD are aiming to bring 4K/60fps gaming to the masses with a wealth of card choices across the midrange and high-end price tiers. Now, if only you can find them for list price!

Note: You can interpret an endorsement for a given GPU (say, the Radeon RX 6700 XT, or the GeForce RTX 3080) as an endorsement for any card based on that GPU, as most should perform within striking distance of one another.

How To Check GPU Health

It isn’t always easy to see how a GPU performs by looking at it with the naked eye. Lots of little things act as clues to indicate whether a graphics card is performing as it should be. Its performance and temperatures are two good ways to determine whether it’s as healthy as you would prefer.

Are There Tests to See How a GPU Is Working

Yes, there are tests you can run with your GPU to see whether it performs as expected. These tests put the entire card under stress by making it do various jobs and then reporting back to you how it’s performing and how it compares to other similar GPUs, depending on which program you use. However, these tests will only tell you how the card is doing and not whether it’s healthy.

What Is the Point of Checking GPU Health

Understanding how healthy your GPU is can help you decide whether you want to purchase a new one or keep your old one working for a while. You can get an idea of whether it’s going to keep working or if you can expect it to die relatively soon. This is especially valuable because you can determine whether you have more time to save and select a new card.

It’s also an essential skill for those who want to purchase a used graphics card. Knowing how to check whether it’s opening correctly and stable will help you know whether it’s worth the purchase. You should never purchase a graphics card without the opportunity to test it first.

What Is a Healthy GPU?

A healthy GPU should meet several different requirements. You don’t only want it to run well and do its job, but you also want to keep it in great shape, so it works for a long time.

You want the GPU to be free of any physical defects. Consider the fans, cables, and heatsink when examining it to see if anything is warped or damaged. Thermal paste or pads should be in good condition or replaced to help the GPU remove heat more effectively.
The GPU should perform up to its expected standard. Of course, a GTX 690 won’t perform as well as a GTX 3090 in the same games — but both should act as you’d expect from a card of that model. You can look up tests to better understand what your GPU can do.
The GPU shouldn’t be overheating or riding the line of nearly too much heat. When a GPU is old or damaged, an early warning sign is getting hotter than it should when using specific programs. Having a good baseline for how hot your GPU usually gets can help you detect issues as soon as they arrive.

Each of these components is important when determining GPU health. Once you’ve run a few stress tests to ensure it works, you should double back and check each of these concerns to ensure your GPU is up to spec.

How to Test GPU Health With Benchmark Testing

One great way to see how a GPU performs is with benchmark testing. Many programs will test the other components in your computer, like the CPU and RAM.

One of the most complex parts is finding a benchmark program that will work for you. If you aren’t using them often, you probably don’t need to pay for a premium program. Look for free programs and free trials. That way, you can run multiple programs and test to see how well the GPU works in each of them.

Choose a benchmark program to use. There are plenty of programs to choose from, and most will work as well as others. Check out companies like Passmark, 3DMark, and Basemark.
Install the program. You want to install it on the computer you’ll be testing.
Follow the instructions to run the program.
Look at the readout. The tests should tell you what percentage your GPU is in, how it compares to other GPUs, and how your computer performs similar to computers with the same components.

Remember, these tests only work if your GPU is active and functional. If you have integrated graphics, switch the GPU you want to benchmark to active.

If your computer and GPU perform similar to those with similar setups, you can determine whether it’s working as it should. This is one of the best GPU health tests you can perform because performance is what matters when it comes to a graphics card.

How to Check a GPU for Physical Defects

Another way to see whether your GPU is working is to examine it for physical defects. The cables should be in good working order. If they’re frayed or burned, they might create issues for the card. If any of the visible pieces of the card are warped, that could make it run less efficiently and stop it from maintaining the heat or power correctly.

One of the most important things to look for is whether the fans are working correctly. If they can’t spin freely, they aren’t cooling the card the way you need them to. Keeping the temperature of your card stable is one of the best ways to maintain its health over the long term.

How to Monitor the Temperature of Your GPU

Monitoring the temperature of your GPU is an intelligent way to ensure you catch problems before they start. Each card manufacturer should have a program that lets you monitor GPU temperatures; there are also third-party programs you can download.

Run the temperature monitors when you do the benchmarking and stress tests. However, you also want to keep them up when you’re doing basic computing,

What if My GPU Doesn’t Pass the Tests?

There are two considerations if your GPU doesn’t pass the test. First, ask yourself which component of the test it failed. Next, determine whether that is a problem that you can fix.

If the card has a physical flaw, like a warped heatsink or frayed cable, check whether you can replace that part. Fans can be swapped, and thermal paste reapplied if needed.

If the card isn’t performing up to your expected standards, remember that not every test is best for every card. Try a few testing services to see how it performs across multiple. Next, decide whether it will still work for your uses. You don’t necessarily need the full power of a graphics card that’s brand new if you’re only doing everyday computing – but you might if you’re playing more recent games on graphics.

If the card is heating up too much, try lowering your graphics settings, cleaning the card, increasing the airflow in your case, or adding an AIO cooler to the GPU. These can help reduce the card’s temperature by providing more cooling than a standard cooling setup with the fans attached to the card.

Of course, if your card is under warranty, contact the company before troubleshooting on your own. Messing with your GPU could void the warranty, and it’s better to go to the company that made it if they’re still obligated to help you.

If you can’t fix your graphics card’s issue, it might be time to seek professional help or even start shopping for a new GPU. Unfortunately, graphics cards rarely fix themselves. Many home remedies — like putting it in the oven — are dicey at best and shouldn’t be attempted unless you don’t mind ruining your card altogether.

Frequently Asked Questions

How Do I Know if My GPU Is Dying?

You might notice it making weird sounds as it performs as if the fans are cranking or groaning. It might have trouble displaying graphics at its average level, leading to a performance loss in games. The temperatures will often start going higher too. Your computer shutting down unexpectedly or games shutting down is another way you might notice your GPU beginning to die.

What Is the Lifespan of a GPU?

You can expect a GPU to function for years. Some last longer than a decade with gentle use. The lifespan is more likely to be an issue in cards that have been used hard or run hot. Usually, cards won’t work as well on newer programs long before they just shut down and die.

What Kills a GPU?

Hard use is the thing most likely to kill a GPU. If you overclock it, the extra voltage can also contribute to it dying sooner than later. If you want to overclock your GPU, only do it as much as you need to. It’s also important to pay attention to temperature because even a GPU used carefully can break with too much heat.

6 Free Programs to Check Your Video Card Memory For Errors • Raymond.CC

If your system memory is faulty, it can cause all manner of weird and wonderful problems, many of which you wouldn’t relate to system RAM being the culprit. Another part of your system that might have its own memory is the video card. That could also fail or become faulty as well. While a problem with system RAM can sometimes be difficult to pin down, video memory problems should be easier to identify because the issues will be related to displaying the image on the screen.

This can manifest itself in a few ways, such as getting a blue screen of death or a restart while playing games. Other issues could be graphical glitches or visual artifacts during gaming. Problems are more likely to occur when your graphics card is working hard. If you suspect there is a fault with your video card’s memory it would be a good idea to test it and help determine if you need to replace or return your card.

Here are six free tools that can help you test the memory of your video card. They were tested with and work in Windows 10 and 7.

1. OCCT

OCCT is a well known testing tool that is able to stress multiple parts of the system like CPU, video card and power supply. It does also have a dedicated option specifically for stress testing video card memory. There is a more general GPU stress testing feature available for testing other areas of video card stability. The OCCT memory test is OpenCL based and can run on pretty much any video card.

After running the program, click on Memtest neat the bottom. Select the video card from the drop down if you have more than one card in the system, then press the big red play button. If any errors are detected, the program will log and display them. OCCT is portable so it doesn’t need to be installed.

Download OCCT

2. GpuMemTest

GpuMemTest is a simple tool to run a number of tests on your video card’s memory. It aims to put stress on the VRAM as well as the memory controller by running a number of test patterns. The tests include sequential, random, alternating read and write, block copy, random data, and sparse inversions.

After install, you only really need to run GpuMemTest and press the Run Test button. The drop down boxes don’t really need touching unless you have a multi GPU set up. A total of eight tests are performed with four to fourteen passes each. Although the website suggests this is only for NVidia video cards, GpuMemTest should work with AMD cards as well because it uses OpenCL (Open Computing Language) for testing.

Download GpuMemTest

3. FurMark

While FurMark doesn’t specifically scan for memory errors, it does run a demanding test that will stress the video card and its memory. Because of that, it could show errors that the video card’s RAM produces. FurMark has been around for many years and is a popular tool to stress test the graphics card for stability when it’s being overclocked. It’s an OpenGL based test and should run on most video cards.

After downloading and installing FurMark, run the program, select the test resolution you want and then Press GPU stress test. Run the test for several minutes and watch for any artifacts during the testing. The very useful GPU-Z and GPU Shark video card information tools are also included.

Download FurMark

4. MSI Kombustor / EVGA OC Scanner X / FurMark Asus ROG Edition

Besides the standard FurMark, there are three other versions made specifically for video card manufacturers MSI, EVGA, and Asus. These versions are customized and have a few differences from the original FurMark. They each have their own user interface skins, some of the tests have company branding, they include Vulkan based tests but they don’t include the GPU-Z tool (EVGA OC Scanner has the NV-Z tool from 2014).

The ROG edition has a useful GLmem tool to check video card memory usage and a floating hardware monitor window. All three programs have custom tests where you can force them to use more VRAM than normal. The MSI and Asus tools can run a test that uses up to an extra 6500MB so most VRAM of an 8GB card can be tested. EVGA OC Scanner can only add up to 3072MB of VRAM (enough for up to 4GB cards).

Enabling the Artifact Scanner should help make it easier to pick up errors because they will be automatically detected. The MSI and Asus versions of FurMark are kept up to date while EVGA OC Scanner X dates back to 2014. It also requires a free EVGA.com account to download the tool. A shared login service like BugMeNot has a few logins for EVGA.com that might help.

Download FurMark Asus ROG Edition | MSI Kombustor | EVGA OC Scanner X

5. Video Memory stress Test

Video Memory stress Test is from back in 2008 and by the same developer behind the useful Check Flash USB drive testing utility. This tool allows you to test your video card RAM using one of three different test types; DirectX, CUDA and OpenGL. It looks similar to the system memory tester Memtest86+ and works in a similar way by running a number of predefined patterns to test the memory as thoroughly as possible.

Three testing sets are present which you can run, they are Full, Express 15% or Small which is good for quickly checking of overall memory integrity. A notable issue is the problems scanning on cards with 3GB or more of dedicated memory. It’s also mentioned online that the CD/floppy version can only read 512MB of video card memory.

It may also fail when trying run tests in 24-bit or 32-bit color display modes, run the vmt.loader.bat in the folder to configure your own set of tests. Because of its age, Video Memory stress Test is probably best suited to older video cards or lower end cards with 1GB/2GB of VRAM.

Download Video Memory Stress Test | Download CD/floppy version

6. MemtestG80/MemTestCL

These two utilities have been grouped together because they have subtle differences but are essentially the same. MemtestG80 will test Nvidia graphics cards that are CUDA enabled (2007 and newer) while MemtestCL tests cards that support OpenCL. Nvidia, AMD, Intel, and any other video cards should be compatible with MemtestCL.

Both versions of Memtest are command line based so to configure them from the defaults you will need to add arguments in Command Prompt or a batch file. Running the executable without arguments will perform a default test of scanning the first 128MB on the first card in the system, and run 50 passes. The tests are a mixture of custom patterns and some based on Memtest86 test patterns.

To test more memory and/or alter the number of times it runs the test, run the program in Command Prompt and add an argument to the command line:

MemtestCL/MemtestG80 [-gpu #] [amount of GPU RAM to test in MB] [number of tests]

If you have more than one graphics card in the system, use the -gpu option, otherwise, it can be excluded. By default, these tools will ask to send statistics to the Folding@Home group, you can disable this option through the command line or press n to answer no. An open source version of MemtestCL is available on Github that excludes the data logging options.

Download MemtestG80 | Download MemtestCL

The page you requested was not found on our site.

A list of future…

List of upcoming PC releases from Sony Interactive Entertainment
LG Display promises to start production…

LG Display promises to start production of 27-inch W-OLED panels this month
TechPowerUp GPU-Z utility released version…

TechPowerUp GPU-Z utility v2.50.0 released with support for NVIDIA GeForce RTX 4090 and Intel Arc 7
PlayStation fans are asking for a remake of The…

PlayStation fans are asking for a remake of The Legend of Dragoon
Horizon Zero remake reportedly.

..

PS5 remake of Horizon Zero Dawn reportedly in development
There are hints of a God release…

There are hints of a God of War Ragnarok release for PC
HWiNFO Gets Early Family Support…

HWiNFO Gets Early Support for AMD’s Zen5 Processor Family
Pictured is the XFX Radeon RX…

Pictured is an XFX Radeon RX 6700 XL with 10 GB of VRAM and a stripped-down Navi 22 GPU.
NVIDIA updated GPU performance.

..

NVIDIA Upgrades Hopper h200 Data Center FP32 GPU Performance from 60 to 67 TFLOPS
ASRock motherboard specs leaked…

ASRock AMD B650, B650E Taichi Motherboard Specifications Leaked — E-ATX with 27-Phase VRM
Motherboard Intel Core i9-13900K…

Motherboard Intel Core i9-13900K and premium Z790 now certified for DDR5-7600 XMP 3.0 memory
Intel Arc A770 GPUs…

Intel Arc A770 and A750 GPUs tested in OpenCL and Vulkan benchmarks

You requested https://gamegpu. com/%25d1%2582%25d0%25b5%25d1%2581%25d1%2582-gpu/action-/-fps-/-tps/ , but no matter how hard our servers try, we couldn’t find it. What happened?

the link you clicked to come here contains a typo
or this page was somehow either deleted or renamed by us
or, which is of course unlikely, did you enter it manually and in doing so made a small mistake?

However, the world does not end there: you might be interested in the following pages on our website:

action-/-fps-/-tps/1
action-/-fps-/-tps
action-/-fps-/-tps/doom-test-gpu
action-/-fps-/ -tps/dark-test-gpu
action-/-fps-/-tps/dayz-test-gpu

The page you requested was not found on our site.

A list of future…

List of upcoming PC releases from Sony Interactive Entertainment
LG Display promises to start production.

..

LG Display promises to start production of 27-inch W-OLED panels this month
TechPowerUp GPU-Z utility released version…

TechPowerUp GPU-Z utility v2.50.0 released with support for NVIDIA GeForce RTX 4090 and Intel Arc 7
PlayStation fans are asking for a remake of The…

PlayStation fans are asking for a remake of The Legend of Dragoon
Horizon Zero remake reportedly…

PS5 remake of Horizon Zero Dawn reportedly in development
There are hints of a God release.

..

There are hints of a God of War Ragnarok release for PC
HWiNFO Gets Early Family Support…

HWiNFO Gets Early Support for AMD’s Zen5 Processor Family
Pictured is the XFX Radeon RX…

Pictured is an XFX Radeon RX 6700 XL with 10 GB of VRAM and a stripped-down Navi 22 GPU.
NVIDIA updated GPU performance…

NVIDIA Upgrades Hopper h200 Data Center FP32 GPU Performance from 60 to 67 TFLOPS
ASRock motherboard specs leaked…

ASRock AMD B650, B650E Taichi Motherboard Specifications Leaked — E-ATX with 27-Phase VRM
Motherboard Intel Core i9-13900K…

Intel Core i9-13900K and premium Z790 motherboard now certified for DDR5-7600 XMP 3.

0 memory
Intel Arc A770 GPUs…

Intel Arc A770 and A750 GPUs tested in OpenCL and Vulkan benchmarks

You requested https://gamegpu.com/mmorpg-/-%25d0%25be%25d0%25bd%25d0%25bb%25d0%25b0%25d0%25b9%25d0%25bd-%25d0%25b8%25d0%25b3% 25d1%2580%25d1%258b/the-cycle-frontier-test-gpu-cpu , but no matter how hard our servers tried, we couldn’t find it. What happened?

the link you clicked to come here contains a typo
or this page was somehow either deleted or renamed by us
or, which is of course unlikely, did you enter it manually and in doing so made a small mistake?

However, the world does not end there: you might be interested in the following pages on our website:

action-/-fps-/-tps/dishonored-death-of-the-outsider-test-gpu-cpu
action-/-fps-/-tps/gothic-playable-teaser-test-gpu-cpu
action-/-fps-/-tps/chernobylite-test-gpu-cpu
action-/-fps-/-tps/chernobylite-test-gpu-cpu-2
action-/-fps-/-tps/grand-theft-auto-the-trilogy-the-definitive-edition-test-gpu-cpu

2022 AMD Graphics Card Ranking

#1	Radeon RX 6950 XT	Type desktop	Performance95. 91	ArchitectureNavi / RDNA2	Year of manufacture2022	Price now1085 USD	TDP335 W
#2	Radeon RX 6900 XT	Type desktop	Capacity87.92	ArchitectureNavi / RDNA2	Year of manufacture2020	Price now940 USD	TDP300 W
#3	Radeon RX 6800 XT	Type Desktop	Capacity81.03	ArchitectureNavi / RDNA2	Year of manufacture2020	Price now659 USD	TDP300 W
#4	Radeon RX 6850M XT	Notebook type	Capacity72.07	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP145 W
#5	Radeon RX 6800	Type Desktop	Productivity71. 20	ArchitectureNavi / RDNA2	Year of manufacture2020	Price now612 USD	TDP250 W
#6	Radeon RX 6750 XT	Type Desktop	Capacity66.18	ArchitectureNavi / RDNA2	Year of manufacture2022	Price now567 USD	TDP250 W
#7	Radeon RX 6700	Type Desktop	Capacity65.94	ArchitectureRDNA 2.0	Year of manufacture2021	Price now567 USD	TDP175 W
#8	Radeon Pro W6800	Workstation Type	Performance65.03	ArchitectureNavi / RDNA2	Year of issue2021	Price now2147 USD	TDP250 W
#9	Radeon RX 6700 XT	Type Desktop	Capacity64. 56	ArchitectureNavi / RDNA2	Year of manufacture2021	Price now424 USD	TDP230 W
#10	Radeon RX 6650 XT	Type Desktop	Performance57.86	ArchitectureNavi / RDNA2	Year of manufacture2022	Price now380 USD	TDP176 W
#11	Radeon RX 5700 XT	Type Desktop	Performance57.23	ArchitectureNavi / RDNA	Year of manufacture2019	Price now245 USD	TDP225 W
#12	Radeon Pro VII	Workstation Type	Performance57.06	ArchitectureGCN 5.1	Year of manufacture2020	Price now2419 USD	TDP250 W
#13	Radeon VII	Type Desktop	Performance56. 99	ArchitectureVega	Year of manufacture2019	Price now402 USD	TDP295 W
#14	Radeon RX 5700 XT 50th Anniversary	Type Desktop	Capacity56.25	ArchitectureRDNA 1.0	Year of manufacture2019	Price now409 USD	TDP225 W
#15	Radeon RX 6600 XT	Type desktop	Productivity54.00	ArchitectureNavi / RDNA2	Year of manufacture2021	Price now347 USD	TDP160 W
#16	Radeon Pro W5700	Workstation type	Performance50.74	ArchitectureRDNA 1.0	Year of manufacture2019	Price now1007 USD	TDP205 W
#17	Radeon Pro Vega II	Workstation Type	Performance49. 90	ArchitectureGCN 5.1	Year of manufacture2019	Price now9999 USD	TDP475 W
#18	Radeon RX Vega 64	Type Desktop	Productivity49.60	ArchitectureVega	Year of manufacture2017	Price now444 USD	TDP295 W
#19	Radeon Pro Vega 64X	Workstation Type	Performance49.54	ArchitectureGCN 5.0	Year of manufacture2019	Price now366 USD	TDP250 W
#20	Radeon RX 5700	Type Desktop	Performance49.51	ArchitectureNavi / RDNA	Year of manufacture2019	Price now261 USD	TDP180 W
#21	Radeon RX 6600	Type Desktop	Performance48. 66	ArchitectureNavi / RDNA2	Year of issue2021	Price now293 USD	TDP132 W
#22	Radeon Pro W6600	Workstation type	Performance48.48	ArchitectureRDNA 2.0	Year of manufacture2021	Price now2077 USD	TDP100 W
#23	Radeon Pro WX 8200	Workstation Type	Productivity47.00	ArchitectureGCN 5.0	Year of manufacture2018	Price now2963 USD	TDP230 W
#24	Radeon RX 5600 XT	Type Desktop	Performance46.82	ArchitectureRDNA 1.0	Year of manufacture2020	Price now436 USD	TDP150W
#25	Radeon RX Vega 56	Type desktop	Capacity46. 15	ArchitectureVega	Year of manufacture2017	Price now350 USD	TDP210 W
#26	Radeon RX 6800S	Notebook Type	Performance46.13	ArchitectureRDNA 2	Year of manufacture2021	Price now −	TDP100 W
№27	Radeon Vega Frontier Edition	Type desktop	Performance45.56	ArchitectureGCN 5.0	Year of manufacture2017	Price now961 USD	TDP300 W
#28	Radeon PRO WX 9100	Type for workstations	Performance45.11	ArchitectureGCN 5.0	Year of manufacture2017	Price now2156 USD	TDP230 W
#29	Radeon RX 6700M	Notebook type	Performance43. 69	ArchitectureRDNA 2	Year of manufacture2020	Price now −	TDP135 W
#30	Radeon RX 6700S	Notebook type	Performance43.55	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP80 W
#31	Radeon Pro 5700 XT	Workstation type	Productivity42.73	ArchitectureRDNA 1.0	Year of manufacture2020	Price now −	TDP130 W
#32	Radeon Pro Vega 56	Workstation Type	Capacity41.98	ArchitectureVega	Year of manufacture2017	Price now4999 USD	TDP210 W
#33	Radeon Pro Vega 64	Type for workstations	Capacity41. 62	ArchitectureGCN 5.0	Year of manufacture2017	Price now6074 USD	TDP250 W
#34	Radeon RX 6650M	Notebook type	Capacity40.33	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP120 W
#35	Radeon Pro 5700	Workstation Type	Performance40.10	ArchitectureRDNA 1.0	Year of manufacture2020	Price now708 USD	TDP130 W
#36	Radeon RX 6600M	Notebook type	Productivity39.34	ArchitectureRDNA 2	Year of manufacture2020	Price now −	TDP100 W
#37	Radeon Pro Vega 48	Workstation type	Capacity38. 44	ArchitectureGCN 5.0	Year of manufacture2019	Price now671 USD	TDP−
#38	Radeon RX 6800M	Type for laptops	Capacity35.84	ArchitectureRDNA 2	Year of manufacture2021	Price now −	TDP145 W
#39	Radeon Pro SSG	Type for workstations	Capacity35.54	ArchitectureGCN 5.0	Year of manufacture2017	Price now821 USD	TDP350 W
#40	Radeon R9 FURY X	Type Desktop	Performance34.31	ArchitectureGCN 3.0	Year of manufacture2015	Price now519 USD	TDP275 W
#41	Radeon RX 590	Type Desktop	Capacity32. 70	ArchitecturePolaris	Year of manufacture2018	Price now113 USD	TDP175 W
#42	Radeon R9 Fury	Type Desktop	Capacity32.43	ArchitectureGCN 1.2	Year of manufacture2015	Price now95 USD	TDP275 W
#43	Radeon Graphics (Ryzen 7000)	Type for laptops	Performance32.17	Architecture	Year of manufacture2022	Price now −	TDP−
#44	Radeon R9 390X	Type Desktop	Productivity32.03	ArchitectureGCN	Year of manufacture2015	Price now506 USD	TDP275W
#45	Radeon Pro W5500	Workstation type	Capacity31. 46	ArchitectureRDNA 1.0	Year of manufacture2020	Price now501 USD	TDP125 W
#46	Radeon RX 6500 XT	Type Desktop	Capacity31.34	ArchitectureNavi / RDNA2	Year of manufacture2022	Price now191 USD	TDP107 W
#47	Radeon RX 5500 XT	Type Desktop	Productivity31.11	ArchitectureRDNA 1.0	Year of manufacture2019	Price now424 USD	TDP130 W
#48	Radeon Pro 5600M	Type Mobile Workstation	Performance30.99	ArchitectureNavi / RDNA	Year of manufacture2020	Price now −	TDP50 W
#49	Radeon R9 390	Type Desktop	Capacity30. 60	ArchitectureGCN 2.0	Year of manufacture2015	Price now380 USD	TDP300W
#50	Radeon RX 580	Type desktop	Capacity30.21	ArchitecturePolaris	Year of manufacture2017	Price now211 USD	TDP185 W
#51	Radeon RX 5500	Type Desktop	Performance29.91	ArchitectureRDNA 1.0	Year of manufacture2019	Price now549 USD	TDP110 W
#52	Radeon RX 480	Type Desktop	Performance29.42	ArchitecturePolaris	Year of manufacture2016	Price now16 USD	TDP150 W
#53	Radeon RX 580 Mobile	Type for laptops	Productivity29. 40	ArchitecturePolaris	Year of manufacture2017	Price now410 USD	TDP100 W
#54	Radeon R9 295X2	Type Desktop	Productivity29.11	ArchitectureGCN 2.0	Year of manufacture2014	Price now600 USD	TDP500 W
#55	Radeon R9 Nano	Type Desktop	Performance28.73	ArchitectureGCN 1.2	Year of manufacture2015	Price now472 USD	TDP175 W
#56	Radeon Pro Duo	Workstation Type	Productivity28.18	ArchitectureGCN 3.0	Year of manufacture2016	Price now4903 USD	TDP350 W
#57	Radeon Pro 5500 XT	Workstation Type	Productivity28. 10	ArchitectureRDNA 1.0	Year of manufacture2020	Price now2499 USD	TDP125 W
#58	Radeon HD 8970M Crossfire	Type for laptops	Performance27.49	ArchitectureGCN	Year of manufacture2012	Price now170 USD	TDP200 W
#59	Radeon RX 470	Type Desktop	Performance27.41	ArchitecturePolaris	Year of release2016	Price now15 USD	TDP120 W
#60	Radeon R9 290	Type Desktop	Productivity27.40	ArchitectureGCN 2.0	Year of manufacture2013	Price now112 USD	TDP275 W
#61	Radeon RX 580 2048SP	Type Desktop	Productivity26. 85	ArchitectureGCN 4.0	Year of manufacture2018	Price now1373 USD	TDP150 W
#62	Radeon Pro WX 7100	Workstation Type	Capacity26.82	ArchitecturePolaris	Year of manufacture2017	Price now582 USD	TDP130 W
#63	Radeon RX 6500M	Type for Notebook	Performance26.71	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP50 W
#64	Radeon RX 580X	Type Desktop	Productivity26.40	ArchitectureGCN 4.0	Year of manufacture2018	Price now191 USD	TDP185 W
#65	Radeon Pro 580	Workstation type	Capacity26. 25	ArchitectureGCN 4.0	Year of manufacture2017	Price now5999 USD	TDP150 W
#66	FirePro W9100	Type for workstations	Performance26.14	ArchitectureGCN 2.0	Year of manufacture2014	Price now1472 USD	TDP275 W
#67	Radeon RX 5600M	Notebook type	Performance25.72	ArchitectureNavi / RDNA	Year of manufacture2020	Price now525 USD	TDP60 W
#68	Radeon Pro 580X	Workstation type	Performance25.53	ArchitectureGCN 4.0	Year of manufacture2019	Price now5999 USD	TDP150 W
#69	Radeon Pro 5300	Workstation type	Performance25. 36	ArchitectureRDNA 1.0	Year of manufacture2020	Price now3649 USD	TDP85 W
#70	Radeon R9 290X	Type Desktop	Performance25.14	ArchitectureGCN	Year of manufacture2013	Price now29 USD	TDP350 W
#71	FirePro W8100	Workstation type	Performance24.76	ArchitectureGCN 2.0	Year of manufacture2014	Price now631 USD	TDP220 W
#72	Radeon RX 5300	Type Desktop	Productivity24.60	ArchitectureRDNA 1.0	Year of manufacture2020	Price now666 USD	TDP100W
#73	Radeon RX 6400	Type desktop	Performance23. 86	ArchitectureRDNA 2.0	Year of manufacture2022	Price now227 USD	TDP53 W
#74	Radeon RX 570	Type Desktop	Productivity23.59	ArchitecturePolaris	Year of manufacture2017	Price now479 USD	TDP120 W
#75	Radeon Pro 5500M	Mobile workstation type	Productivity23.04	ArchitectureNavi / RDNA	Year of manufacture2019	Price now1950 USD	TDP50 W
#76	Radeon R9 M290X Crossfire	Type for laptops	Productivity23.03	ArchitectureGCN	Year of manufacture2014	Price now −	TDP200 W
#77	Radeon RX Vega M GH	Notebook type	Performance22. 87	ArchitectureVega	Year of manufacture2018	Price now1031 USD	TDP100 W
#78	Radeon RX Vega M GL / 870	Notebook type	Performance22.77	ArchitectureVega	Year of manufacture2018	Price now −	TDP65 W
#79	Radeon R9 285	Type Desktop	Performance22.62	ArchitectureGCN 3.0	Year of manufacture2014	Price now73 USD	TDP190 W
#80	Radeon Pro 570	Workstation type	Performance21.45	ArchitectureGCN 4.0	Year of manufacture2017	Price now157 USD	TDP120 W
#81	Radeon HD 7970M Crossfire	Type for laptops	Capacity21. 44	ArchitectureGCN	Year of manufacture2012	Price now126 USD	TDP200 W
#82	Radeon R9 380	Type Desktop	Performance21.19	ArchitectureGCN	Year of manufacture2015	Price now163 USD	TDP220 W
#83	Radeon R9 380X	Type Desktop	Performance21.01	ArchitectureGCN 3.0	Year of manufacture2015	Price now17 USD	TDP250 W
#84	Radeon Pro W6400	Workstation Type	Productivity20.91	ArchitectureRDNA 2.0	Year of manufacture2022	Price now256 USD	TDP50 W
#85	FirePro W9000	Type for workstations	Performance20. 78	ArchitectureGCN 1.0	Year of manufacture2012	Price now913 USD	TDP274 W
#86	Radeon Pro WX Vega M GL	Type for mobile workstations	Performance20.48	ArchitectureVega	Year of manufacture2018	Price now1359 USD	TDP65 W
#87	Radeon Pro 5300M	Mobile workstation type	Performance19.82	ArchitectureNavi / RDNA	Year of release2019	Price now2068 USD	TDP65 W
#88	Radeon R9 280X	Type Desktop	Performance19.76	ArchitectureGCN	Year of manufacture2013	Price now190 USD	TDP200 W
#89	FirePro W7100	Type for workstations	Productivity19. 62	ArchitectureGCN 3.0	Year of manufacture2014	Price now414 USD	TDP150 W
#90	Radeon R9 280	Type Desktop	Performance19.15	ArchitectureGCN 1.0	Year of manufacture2014	Price now228 USD	TDP200 W
#91	Radeon RX 470 Mobile	Type for laptops	Performance18.87	ArchitecturePolaris	Year of manufacture2016	Price now148 USD	TDP120 W
#92	Radeon HD 7990	Type Desktop	Performance18.84	ArchitectureGCN 1.0	Year of manufacture2013	Price now537 USD	TDP375 W
#93	Radeon RX Vega 8 (Ryzen 5000)	Notebook type	Performance18. 75	ArchitectureVega	Year of manufacture2020	Price now −	TDP15 W
#94	Radeon Pro WX 5100	Workstation type	Productivity18.55	ArchitectureGCN 4.0	Year of manufacture2016	Price now1498 USD	TDP75 W
#95	Radeon 680M	Notebook Type	Performance18.32	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP45 W
#96	FirePro S10000	Type for workstations	Productivity18.00	ArchitectureGCN 1.0	Year of manufacture2012	Price now699 USD	TDP375 W
#97	Radeon RX Vega 8 (Ryzen 4000/5000)	Notebook type	Performance17. 93	ArchitectureVega	Year of release2020	Price now −	TDP15 W
#98	Radeon HD 7970	Type Desktop	Performance17.77	ArchitectureGCN	Year of manufacture2011	Price now233 USD	TDP250 W
#99	Radeon R9 M295X	Notebook type	Productivity17.44	ArchitectureGCN 3	Year of manufacture2014	Price now −	TDP125 W
#100	Radeon R9 M395	Notebook type	Performance17.40	ArchitectureGCN	Year of manufacture2015	Price now1288 USD	TDP−
#101	Radeon Pro Vega 20	Type for mobile workstations	Performance17. 32	ArchitectureVega	Year of manufacture2018	Price now360 USD	TDP100 W
#102	Radeon RX 560X (Mobile)	Notebook type	Performance17.25	ArchitecturePolaris	Year of release2017	Price now520 USD	TDP60 W
#103	FirePro S9000	Workstation type	Performance17.13	ArchitectureGCN 1.0	Year of manufacture2012	Price now400 USD	TDP225 W
#104	Radeon R9 270X	Type Desktop	Performance16.67	ArchitectureGCN	Year of manufacture2013	Price now229 USD	TDP180 W
#105	FirePro S9050	Workstation type	Performance16. 59	ArchitectureGCN 1.0	Year of manufacture2014	Price now550 USD	TDP225 W
#106	Radeon R9 M395X	Notebook type	Performance16.17	ArchitectureGCN 3	Year of manufacture2014	Price now −	TDP125 W
#107	Radeon HD 7950	Type Desktop	Performance16.13	ArchitectureGCN 1.0	Year of issue2012	Price now348 USD	TDP200 W
#108	Radeon Sky 500	Type Desktop	Performance16.04	ArchitectureGCN 1.0	Year of manufacture2013	Price now129 USD	TDP150 W
#109	Radeon R9 370	Type Desktop	Performance15. 99	ArchitectureGCN 1.0	Year of manufacture2015	Price now352 USD	TDP110 W
#110	Radeon Pro Vega 16	Mobile workstation type	Performance15.85	ArchitectureVega	Year of manufacture2018	Price now511 USD	TDP75 W
#111	Radeon HD 6990M Crossfire	Type for Notebook	Performance15.85	ArchitectureTerascale 2	Year of manufacture2011	Price now170 USD	TDP−
#112	Radeon HD 7870	Type Desktop	Performance15.80	ArchitectureGCN	Year of issue2012	Price now300 USD	TDP200 W
#113	Radeon RX 5500M	Notebook type	Performance15. 71	ArchitectureNavi / RDNA	Year of manufacture2019	Price now998 USD	TDP85 W
#114	Radeon R7 370	Type Desktop	Performance15.23	ArchitectureGCN 1.0	Year of manufacture2015	Price now283 USD	TDP100 W
#115	Radeon HD 7870 XT	Type Desktop	Performance15.13	ArchitectureGCN 1.0	Year of manufacture2012	Price now133 USD	TDP185 W
#116	Radeon RX 550 (Mobile)	Notebook type	Performance15.12	ArchitecturePolaris	Year of manufacture2017	Price now294 USD	TDP50 W
#117	FirePro W7000	Type for workstations	Performance14. 60	ArchitectureGCN 1.0	Year of manufacture2012	Price now258 USD	TDP150 W
#118	Radeon RX Vega 8 (Ryzen 4000)	Notebook type	Performance14.59	ArchitectureVega	Year of manufacture2020	Price now631 USD	TDP15 W
#119	Radeon R9 270	Type desktop	Performance14.58	ArchitectureGCN 1.0	Year of manufacture2013	Price now237 USD	TDP150 W
#120	FirePro S7000	Workstation type	Performance14.11	ArchitectureGCN 1.0	Year of manufacture2012	Price now257 USD	TDP150 W
#121	Radeon RX 460	Type Desktop	Performance13. 96	ArchitecturePolaris	Year of manufacture2016	Price now427 USD	TDP75 W
#122	Radeon HD 6970M Crossfire	Notebook type	Productivity13.87	ArchitectureTerascale 2	Year of manufacture2011	Price now170 USD	TDP−
#123	FirePro W8000	Type for workstations	Performance13.87	ArchitectureGCN 1.0	Year of manufacture2012	Price now281 USD	TDP225 W
#124	Radeon RX Vega 7	Type for Notebook	Performance13.80	ArchitectureVega	Year of manufacture2020	Price now387 USD	TDP15 W
#125	Radeon HD 8970M	Notebook type	Performance13. 12	ArchitectureGCN	Year of issue2013	Price now135 USD	TDP100 W
#126	Radeon RX 5300M	Notebook Type	Productivity13.05	ArchitectureNavi / RDNA	Year of manufacture2019	Price now525 USD	TDP65 W
#127	Radeon RX 550X (Mobile)	Notebook type	Performance12.97	ArchitecturePolaris	Year of manufacture2019	Price now299 USD	TDP50 W
#128	Radeon HD 7850	Type Desktop	Performance12.96	ArchitectureGCN 1.0	Year of manufacture2012	Price now278 USD	TDP130 W
#129	Radeon HD 7970M	Notebook Type	Performance12. 78	ArchitectureGCN	Year of manufacture2012	Price now338 USD	TDP100 W
#130	Radeon Pro WX 4100	Workstation Type	Performance12.63	ArchitectureGCN 4.0	Year of manufacture2016	Price now1070 USD	TDP50 W
#131	Radeon Pro 560X	Mobile workstation type	Performance12.45	ArchitecturePolaris	Year of manufacture2017	Price now133 USD	TDP35 W
#132	Radeon RX Vega 6 (Ryzen 4000)	Type for laptops	Performance12.44	ArchitectureVega	Year of manufacture2020	Price now717 USD	TDP−
#133	Radeon 660M	Notebook type	Performance12. 26	ArchitectureRDNA 2	Year of manufacture2022	Price now −	TDP45 W
#134	Radeon RX 560	Type Desktop	Performance12.19	ArchitectureGCN 4.0	Year of manufacture2017	Price now355 USD	TDP75 W
#135	Radeon RX Vega 6 (Ryzen 4000/5000)	Notebook type	Performance12.19	ArchitectureVega	Year of manufacture2020	Price now −	TDP15 W
#136	Radeon RX 460 (Mobile)	Notebook type	Performance12.05	ArchitecturePolaris	Year of manufacture2016	Price now95 USD	TDP75 W
#137	Radeon RX 560 Mobile	Type for laptops	Performance11. 88	ArchitecturePolaris	Year of manufacture2017	Price now468 USD	TDP75 W
#138	Radeon Pro 560	Mobile workstation type	Performance11.77	ArchitecturePolaris	Year of release2017	Price now127 USD	TDP35 W
#139	Radeon Pro 460	Type Mobile Workstation	Performance11.69	ArchitecturePolaris	Year of manufacture2016	Price now100 USD	TDP35 W
#140	Mobility Radeon HD 5870 Crossfire	Notebook Type	Capacity 11.63	ArchitectureTerascale 2	Year of manufacture2010	Price now56 USD	TDP120 W
#141	Radeon R9 M470X	Notebook type	Performance11. 51	ArchitectureGCN 2	Year of manufacture2016	Price now −	TDP−
#142	Radeon RX Vega M GL	Notebook type	Performance11.48	ArchitectureGCN 4.0	Year of manufacture2018	Price now1307 USD	TDP65 W
#143	Radeon R9 M290X	Notebook Type	Performance11.10	ArchitectureGCN	Year of manufacture2014	Price now613 USD	TDP100 W
#144	Radeon Pro 555X	Mobile workstation type	Performance10.95	ArchitecturePolaris	Year of manufacture2017	Price now894 USD	TDP−
#145	Radeon R9 M485X	Type for laptops	Performance10. 85	ArchitectureGCN 3	Year of manufacture2016	Price now2500 USD	TDP125 W
#146	Radeon RX 560X	Type Desktop	Performance10.77	ArchitectureGCN 4.0	Year of manufacture2018	Price now1003 USD	TDP75 W
#147	FirePro W7170M	Type for mobile workstations	Performance10.70	ArchitectureGCN 3	Year of manufacture2015	Price now600 USD	TDP100 W
#148	Radeon Pro 555	Type Mobile Workstation	Performance10.63	ArchitecturePolaris	Year of manufacture2017	Price now894 USD	TDP−
#149	Radeon R7 260X	Type Desktop	Performance10. 62	ArchitectureGCN 2.0	Year of manufacture2013	Price now139 USD	TDP115 W
#150	Radeon Pro 455	Type for mobile workstations	Performance10.54	ArchitecturePolaris	Year of manufacture2016	Price now696 USD	TDP−
#151	Radeon HD 7790	Type Desktop	Performance10.46	ArchitectureGCN 2.0	Year of manufacture2013	Price now478 USD	TDP85 W
#152	Radeon R7 360	Type Desktop	Performance10.31	ArchitectureGCN 2.0	Year of manufacture2015	Price now14 USD	TDP80 W
#153	Radeon R9 M390X	Notebook type	Performance10. 23	ArchitectureGCN 3	Year of manufacture2015	Price now −	TDP125 W
#154	FirePro W5100	Type for workstations	Performance10.21	ArchitectureGCN 2.0	Year of manufacture2014	Price now314 USD	TDP50 W
#155	Radeon RX Vega 9	Type for laptops	Performance10.20	ArchitectureVega	Year of manufacture2017	Price now302 USD	TDP15 W
#156	Radeon HD 6990	Type Desktop	Performance10.20	ArchitectureTeraScale 3	Year of manufacture2011	Price now223 USD	TDP375 W
#157	FirePro W5000	Type for workstations	Performance10. 15	ArchitectureGCN 1.0	Year of manufacture2012	Price now110 USD	TDP75 W
#158	Radeon R9 M380	Notebook Type	Productivity9.83	ArchitectureGCN	Year of manufacture2015	Price now −	TDP−
#159	FirePro W4300	Type for workstations	Productivity9.80	ArchitectureGCN 2.0	Year of manufacture2015	Price now229 USD	TDP50 W
#160	Radeon R7 260	Type desktop	Performance9.79	ArchitectureGCN 2.0	Year of manufacture2013	Price now139 USD	TDP95 W
#161	Radeon Pro V340 MxGPU	Workstation Type	Performance9. 66	ArchitectureGCN 5.0	Year of manufacture2018	Price now −	TDP300 W
#162	Radeon HD 6970	Type Desktop	Performance9.57	ArchitectureTerascale 2	Year of manufacture2010	Price now233 USD	TDP250 W
#163	Radeon RX Vega 5	Notebook type	Productivity9.49	ArchitectureVega	Year of manufacture2020	Price now287 USD	TDP15 W
#164	Radeon RX 540	Notebook type	Performance9.44	ArchitecturePolaris	Year of manufacture2017	Price now166 USD	TDP75 W
#165	Radeon RX 480 Mobile	Type for laptops	Performance9. 37	ArchitecturePolaris	Year of manufacture2016	Price now577 USD	TDP100 W
#166	Radeon R9 M390	Notebook type	Performance9.28	ArchitectureGCN	Year of manufacture2015	Price now −	TDP−
#167	FirePro S7150	Type for workstations	Performance9.23	ArchitectureGCN 3.0	Year of manufacture2016	Price now1699 USD	TDP150 W
#168	Radeon Pro 450	Type for mobile workstations	Performance9.22	ArchitecturePolaris	Year of manufacture2016	Price now183 USD	TDP75 W
#169	Radeon RX 550	Type Desktop	Performance9. 15	ArchitectureGCN 4.0	Year of manufacture2017	Price now156 USD	TDP50 W
#170	FirePro R5000	Type for workstations	Performance8.96	ArchitectureGCN 1.0	Year of manufacture2013	Price now500 USD	TDP150 W
#171	Radeon HD 6950	Type Desktop	Performance8.94	ArchitectureTeraScale 3	Year of manufacture2010	Price now191 USD	TDP200 W
#172	Radeon HD 6990M	Notebook type	Performance8.93	ArchitectureTerascale 2	Year of manufacture2011	Price now190 USD	TDP100 W
#173	Radeon PRO WX 3100	Mobile workstation type	Performance8. 83	ArchitecturePolaris	Year of manufacture2017	Price now242 USD	TDP65 W
#174	Radeon HD 6850 X2	Type Desktop	Performance8.58	ArchitectureTeraScale 2	Year of manufacture2011	Price now −	TDP254 W
#175	FireStream 9370	Type desktop	Performance8.56	ArchitectureTeraScale 2	Year of manufacture2010	Price now −	TDP225 W
#176	Radeon RX Vega 8 (Ryzen 2000/3000)	Notebook type	Performance8.54	ArchitectureVega	Year of release2017	Price now −	TDP15 W
#177	Radeon Pro WX 3200	Workstation type	Performance8. 37	ArchitecturePolaris	Year of manufacture2019	Price now740 USD	TDP65 W
#178	Radeon RX 550X	Type Desktop	Performance8.01	ArchitectureGCN 4.0	Year of manufacture2018	Price now259 USD	TDP50 W
#179	FirePro V7900	Type for workstations	Capacity7.80	ArchitectureTeraScale 3	Year of manufacture2011	Price now172 USD	TDP150 W
#180	Radeon R9 M470	Type for laptops	Performance7.75	ArchitectureGCN 2	Year of manufacture2016	Price now −	TDP−
#181	Radeon HD 6970M	Notebook type	Performance7. 69	ArchitectureTerascale 2	Year of manufacture2011	Price now257 USD	TDP75 W
#182	Radeon R7 250X	Type Desktop	Performance7.68	ArchitectureGCN 1.0	Year of manufacture2014	Price now207 USD	TDP80 W
#183	Radeon HD 5870	Type Desktop	Productivity7.68	ArchitectureTeraScale 2	Year of manufacture2009	Price now124 USD	TDP188 W
#184	Radeon HD 5970	Type Desktop	Performance7.59	ArchitectureTeraScale 2	Year of manufacture2009	Price now69 USD	TDP294 W
#185	Radeon RX Vega 6 (Ryzen 2000)	Type for laptops	Performance7. 51	ArchitectureVega	Year of manufacture2018	Price now248 USD	TDP−
#186	Radeon RX Vega 6 (Ryzen 2000/3000)	Notebook Type	Performance7.51	ArchitectureVega	Year of manufacture2018	Price now −	TDP15 W
#187	Radeon HD 6870	Type Desktop	Capacity7.50	ArchitectureTerascale 2	Year of manufacture2010	Price now159 USD	TDP151 W
#188	Radeon Pro WX 4150	Type Mobile Workstation	Performance7.49	ArchitecturePolaris	Year of manufacture2017	Price now2000 USD	TDP50 W
#189	Mobility Radeon HD 4870 X2	Notebook type	Performance7. 48	ArchitectureTerascale 1	Year of manufacture2009	Price now305 USD	TDP−
#190	FirePro W5170M	Type for mobile workstations	Capacity7.42	ArchitectureGCN	Year of manufacture2015	Price now1530 USD	TDP−
#191	Radeon HD 7770	Type Desktop	Capacity7.41	ArchitectureGCN	Year of manufacture2012	Price now315 USD	TDP80 W
#192	Radeon RX Vega 11	Type Desktop	Performance7.19	ArchitectureVega	Year of manufacture2017	Price now672 USD	TDP65 W
#193	Mobility Radeon HD 5870	Notebook Type	Performance7. 13	ArchitectureTerascale 2	Year of manufacture2010	Price now108 USD	TDP50 W
#194	FirePro M5100	Type for mobile workstations	Performance7.12	ArchitectureGCN	Year of manufacture2013	Price now105 USD	TDP−
#195	Radeon R9 M385	Notebook Type	Capacity6.83	ArchitectureGCN	Year of manufacture2015	Price now999 USD	TDP−
#196	Radeon Pro W5500M	Workstation type	Performance6.81	ArchitectureRDNA 1.0	Year of manufacture2020	Price now2381 USD	TDP85 W
#197	Radeon R9 M385X	Notebook Type	Capacity6. 75	ArchitectureGCN	Year of manufacture2015	Price now999 USD	TDP−
#198	Radeon HD 5850	Type desktop	Performance6.73	ArchitectureTerascale 2	Year of manufacture2009	Price now141 USD	TDP151 W
#199	Radeon HD 6850	Type Desktop	Performance6.71	ArchitectureTeraScale 2	Year of manufacture2010	Price now258 USD	TDP127W
#200	Radeon R7 M275DX	Notebook Type	Performance6.71	ArchitectureGCN	Year of manufacture2014	Price now −	TDP−

Video cards | Laboratory | Overclockers.ru

All articles

ASUS

Video cards

Processors

motherboards

Cooling

Periphery

Software

Games

FAQ

Memory

Power supplies

Corps

Everything about. ..

HDD drives

Monitors

Overclocking Institute

network hardware

SSD drives

Mobile devices

Smartphones

Tablets

laptops

Guides

Developments

Website

ATI
options

Radeon 8500-9250

Radeon 9500-9600-9800SE

Radeon 9700-9800

Radeon X300-X600

Radeon X700

Radeon X800-X850

Radeon X1300

Radeon X1600

Radeon X1800

Radeon X1900

Radeon HD 2400

Radeon HD 2600

Radeon HD 2900

Radeon HD 3800

Radeon HD 4800

Radeon HD 4600

Radeon HD 4700

Radeon HD 5800

Radeon HD 5700

Radeon HD 5900

Radeon HD 5600

Radeon HD 5400

Radeon HD 3200

Radeon HD 6800

Radeon HD 6900

Radeon HD 6700

Radeon HD 6600

Radeon HD 7900

Radeon HD 7700

Radeon HD 7800

Radeon R9280X

Radeon R9 270X

Radeon R7 260X

Radeon R9 290/290X

Radeon R9 390X

Radeon R7 370

Radeon R9 390

Radeon R9 Fury X

Radeon R9 380

Radeon R9Fury

Radeon R7 360

Radeon R9 Nano

Radeon R9 380X

Radeon RX 480

Radeon RX 470

Radeon RX 460

Radeon RX 580

Radeon RX 550

Radeon RX 570

Radeon RX Vega 64

Radeon RX 560

Radeon RX Vega 56

Radeon RX 6800

Radeon RX 6700

Radeon RX 6600

NVIDIA
options

GeForce MX

GeForce 3

GeForce 4

GeForceFX

GeForce 6

GeForce 7

GeForce 8

GeForce 9

GeForce GTX

GeForce GTS

GeForce GT

GeForce GTX 680

GeForce GTX 690

GeForce 600

GeForce 700

GeForce 980

GeForce 970

GeForce 960

GeForce GTX Titan X

GeForce 980 Ti

GeForce 950

GeForce GTX 1080

GeForce GTX 1070

GeForce GTX 1060

GeForce GTX 1050 Ti

GeForce GTX 1050

Titan X (Pascal)

GeForce GTX 1080 Ti

GeForce GT 1030

GeForce GTX 1070 Ti

GeForce RTX 2080 Ti

GeForce RTX 2080

GeForce RTX 2070

GeForce RTX 2060

GeForce GTX 1660 Ti

GeForce GTX 1660

GeForce RTX 3080

GeForce RTX 3090

GeForce RTX 3070

GeForce RTX 3060 Ti

GeForce RTX 3060

GeForce RTX 3080 Ti

GeForce RTX 3070 Ti

By manufacturer
options

Abit

Albatron

ASUS

chaintech

Gainward

gigabyte

Leadtek

MSI

Palit

Power Color (Tul)

Prolink

Sapphire

Sparkle

SUMA

Other

HIS

XFX

bfg

Zotac

Sysconn

point of view

Interface
options

AGP

PCI Express

Dmitry Olegovich

June 10, 2022

Testing the Radeon RX 6800 XT video card and five other graphics accelerators at 1920 x 1080, 2560 x 1440, 3840 x 2160 resolutions and two operating modes.

Dmitry Olegovich

June 1, 2022

Testing the Radeon RX 6800 and five other GPUs at 19 resolutions20 x 1080, 2560 x 1440, 3840 x 2160 and two modes of operation.

Dmitry Olegovich

May 27, 2022

Testing the Radeon RX 6700 XT video card and five other graphics accelerators at 1920 x 1080, 2560 x 1440, 3440 x 1440 resolutions and two operating modes.

Dmitry Olegovich

May 22, 2022

Testing the Radeon RX 6600 XT video card and five other graphics accelerators at 1920 x 1080, 2560 x 1440, 3440 x 1440 resolutions and two operating modes.

Dmitry Olegovich

May 18, 2022

Testing the Radeon RX 6600 video card and five other graphics accelerators at 1920 x 1080, 2560 x 1080, 2560 x 1440 resolutions and two operating modes.

Maxim Romanov

May 12, 2022

Choosing a PCI video card for a retro platform based on Windows XP: six generations of models, processor dependence, system bus frequency and other characteristics that will determine the winner.

Dmitry Olegovich

March 27, 2022

Testing two PC configurations in two operating modes, 1920 x 1080 resolution and eight games.

Dmitry Olegovich

March 25, 2022

Testing NVIDIA GeForce RTX 2070 Super and AMD Radeon RX 5700 XT video cards in conjunction with a Core i7-11700KF processor at 1920 x 1080, 2560 x 1080 and 2560 x 1440 resolutions, two operating modes and eight games.

Alexander Anderson

March 24, 2022

The MSI model belongs to that rare tribe of budget graphics accelerators that deserve attention.

Dmitry Olegovich

March 20, 2022

The higher the resolution, the greater the performance degradation. And according to this indicator, we will find out how much the volume affects the performance of graphics accelerators.

Dmitry Olegovich

March 18, 2022

Testing NVIDIA GeForce RTX 2060 Super and AMD Radeon RX 5700 video cards in conjunction with a Core i7-11700F processor with 19 resolutions20 x 1080, 2560 x 1080 and 2560 x 1440, two modes of operation and eight games.

Alexander Anderson

March 10, 2022

The new accelerator is an intermediate link between the two flagships and is designed to close the gap between them not only in terms of performance, but also in price.

Dmitry Olegovich

March 9, 2022

Testing two PC configurations in two operating modes, 1920 x 1080 resolution and eight games.

Andrey Ponkratov

March 8, 2022

New to The Ultimate Force series based on Navi 24 is a great quiet graphics card for online gaming and light gaming that is built to last.

Dmitry Olegovich

March 7, 2022

Test NVIDIA GeForce RTX 3060, GeForce RTX 2060, GeForce GTX 1070 and AMD Radeon RX 5600 XT at 1920 x 1080, 2560 x 1080 and 2560 x 1440 resolutions, two modes and eight games.

Dmitry Olegovich

March 2, 2022

Test 20 (Radeon RX 5500 XT to Radeon RX 5700 XT, GeForce GTX 1650 to GeForce RTX 3080) GPUs in eight games and three resolutions.

Andrey Ponkratov

March 1, 2022

The new generation of NVIDIA takes into account all the features of video cards with GDDR6X memory, including their heating.

Dmitry Olegovich

February 20, 2022

Testing six AMD and NVIDIA graphics accelerators and two processors in eight games, two operating modes and a resolution of 1920 x 1080.

Dmitry Olegovich

February 18, 2022

Testing two PC configurations in two operating modes, 1920 x 1080 resolution and eight games.

Dmitry Olegovich

February 16, 2022

Testing NVIDIA GeForce GTX 1660 Ti and AMD Radeon RX 5600 XT video cards in conjunction with a Core i7-11700KF processor at 1920 x 1080, 2560 x 1080 and 2560 x 1440 resolutions, two operating modes and eight games.

Video card performance test

How to check the video card for its performance? Setting such a goal is usually faced by those who want to buy this device at an affordable price in the used components market. Well, or preparing for sale on the used market. But we will consider the situation more in the context of the first case. There are many arguments against buying used video cards in principle, however, if there is no financial opportunity to purchase this device new, if a familiar person sells it, a service center as a legal entity, or if a minimizing our risks are an agreement scheme, then you can go for the used option. Well, we will now deal with checking the performance of the video card.

Content

How to check the video card for its performance
Stress testing of the video card
Testing in real loads
Testing Furmark

The first stage is an external examination of the device, whether there are any signs of overheating on it, whether the cooler is intact, whether it rotates freely, whether the video outputs are in place, whether they are soldered by rough soldering, whether there are swollen capacitors (if the video card is without a body and allows such an inspection), etc.
The second step is to connect the video card to the computer. It is checked whether the device is working in principle. In order for the computer to see it, in some cases, for this you need to reset the BIOS settings to default.
The third step is to install the video driver in the Windows environment and visually check if everything is in order with color correction, if there are artifacts on the screen, if the video driver fails, if there are any lags when playing video, minimizing and full-screen windows maximizing, etc.
The fourth step is to check the temperature in idle, without loads. For graphics cards with active cooling (with coolers), the normal idle temperature ranges from 30°C to 44°C depending on the model. For passively cooled devices, 50-55°C can be normal when idle.
The fifth stage is stress testing of the video card, tracking temperature fluctuations and fixing the reached temperature limit. But let’s talk about the stage in more detail.

Video card stress testing

Video card stress testing is testing it with special programs under maximum loads. Special programs simulate tasks by fully loading the video card so that we can see how it behaves at maximum loads. How the temperature fluctuates, to what boundary value it rises and stabilizes or does not stabilize at all. Will this value be greater than the temperature limit declared by the manufacturer for this particular video card model. Will artifacts, stripes, ripples, color correction bugs appear on the screen during stress testing. Will the video driver crash, crashing.

Stress testing is carried out by special programs, they offer tests with different loads. There are tests with smaller loads, as, for example, in the AIDA64 program. And there are tough tests that put video cards under real stress with the most maximum loads, such as, for example, in the FurMark program. Below we will look at how tests are carried out using these two programs.

But first, a few words about the stress testing process itself. It usually takes 20 minutes to complete. 20 minutes is not just long enough for stress testing, with tough tests like FurMark, testing the device for more than this time is even undesirable. For 20 minutes, you can track the temperature increase and fix either, at best, the temperature stabilizes at some acceptable limit value, or, in the worst case, an excess of the limit value. Stabilization of the temperature curve, reaching its value no higher than the boundary limit declared by the manufacturer, the absence of artifacts and other signs of the problematic operation of the video card mentioned above — all this will mean that the test has been passed successfully.

But stress testing is not necessarily carried out with special programs, you can simply subject the video card to those specific loads that it will constantly face. And, perhaps, to slightly increase these loads.

Testing under real loads

In the case of gaming video cards, we can simply run some game installed on the computer at maximum speed. Naturally, this should be a modern game capable of giving the video card a real load. Thus, we will check the device under conditions of maximum loads, but not those that can be in principle, and which we may never encounter in everyday life, but under conditions of real maximum loads. Testing under real load conditions can be carried out for more than 20 minutes, you can play at maximum speeds for about an hour. And during this hour you need to track the temperature in the background. To monitor the temperature, you can resort to using the GPU-Z program. Run it, go to the «Sensors» tab. Here we have the “GPU Temperature” column, it shows the current temperature of the video card. By the way, please note: on laptops with two video cards, the desired one is selected at the bottom of the program window.

But we need to record the maximum temperature of the video card during the test game. Therefore, in the drop-down list of this column, we indicate the display of the maximum temperature that the GPU-Z program recorded during the current session of its work — we tick the “Show Highest Reading” checkbox.

Thus, taking a break from the game from time to time, we will be able to see the maximum temperature value noted by GPU-Z for the entire time that we played.

FurMark Test

The FurMark program provides the ability to subject the video card to extreme, aggressive loads. It will not be so heavily loaded in its work by either games or programs aimed at working with graphics, as it does during the FurMark test. This is, so to speak, a check with a gap. The FurMark test is called the killer of video cards, since this program, in addition to the usual stress test for checking stability, also provides for an extreme test that can potentially destroy a video card. It is advisable not to experiment with the FurMark settings without the need, it is better to just leave the default testing options.

Before starting the test, select the screen resolution in which we want to test the video card. Click the «GPU stress test» button.

The program warns us that an operation that is intensively loading the GPU is about to be carried out. And what if we have critical overclocking parameters, a power supply of insufficient power, or there are problems with cooling, then we ourselves bear all the risks of running the test. If these problems do not exist, click «Go».

A picture of a hairy donut will appear on the screen — a 3D image with complex rendering, it is this monster that subjects video cards to the heaviest loads. At the top of the FurMark active test window, we will see the current temperature of the video card. At the bottom of the window, we will observe on the graph a growth curve and, at best, temperature stabilization.

To stop the test, simply close the hairy donut window.

Note: is not recommended for use with FurMark for passively cooled graphics cards. These video cards are not designed for heavy loads, their cooling system is designed to provide office, maximum multimedia tasks. And in principle, there is no point in putting any video cards at risk if they do not work at the strain of their capabilities. To check the stability of their work, the usual stress test offered by AIDA64 will be quite enough.

AIDA64 testing

The AIDA64 stress test is the best way to test all non-gaming video cards, they will simply be subjected to simulated background loads. In the main window of the program, we launch a general stability test, you can use the button on the toolbar.

If we are working with the old version of AIDA64, and there is no such button, then in the «Service» menu, click the «System stability test» item. Further, in the block at the top left, we remove the options from all the hardware components of the computer except for the video card. By unchecking the display of all curves on the temperature graph so that the curves of other devices do not interfere with us. Click the «Preferences» button.

And here we add the display of the GPU Diode curve with the color you like, for example, with red.

Start the test with the «Start» button.

And we monitor the temperature curve of the video card on the first graph. At the end of the graph, we will see the current temperature value.

To stop the test, press the «Stop» button.

How to check a video card on a computer

How to check a video card — Checking a video card for performance — Quick and easy!

How to check the video card. Checking the video card directly in the mail

How to check a video card without connecting to a PC

choosing the best of the nine new GPUs / Sudo Null IT News
In the spring of 2021, NVIDIA introduced a new line of RTX Ax000 and Ax0 graphics cards based on the Ampere architecture, with third-generation tensor cores.

At that time, it was already possible to rent dedicated and cloud servers with Tesla M60, T4, V100 GPUs and even top-end NVIDIA A100s from Selectel.

Since we strive to provide our customers with only up-to-date hardware with modern technologies, we decided that it was time to update the line of video cards. Offering all the video cards announced by NVIDIA is irrational both for us and for our customers. Under the cut, I’ll tell you how we chose the best of the best and share the results of our benchmark on a test build.

The approach by which we at Selectel choose hardware — video cards, processors and other components — is quite simple. We assume that the client wants to solve their business problems efficiently and at minimal cost. Accordingly, we start from the following formula:

According to it, we chose leaders among new video cards.

Which video cards were considered

We compared nine GPUs: RTX video cards from A2000 to A6000, A10, A16, A30, A40 and A100 PCIe. A2000 was only released this summer, but that didn’t stop us from reviewing the chip’s specifications and testing the sample.

Here we have a «family» of RTX Ax000 — from the older A5000 to the younger A2000.

All test participants are server video cards, desktop GeForce RTX 3080 and 3090 are not on the list. These cards (and to be precise, the installation of NVIDIA drivers) are prohibited for use in servers in data centers. The manufacturer strictly monitors compliance with the restrictions: sanctions for violation are applied not only to the provider, but also to the client who rents a server with desktop hardware or installs NVIDIA software on it.

To evaluate video cards, we started from several characteristics that are important for solving problems that customers often have. That is, we looked at what, in general, these GPUs are taken for. The purpose of the cores is presented in a simplified form, each type affects the performance of the video card.

Among them:

The number of CUDA cores (for those who don’t know, this is a symbol for scalar computing units in NVIDIA video chips). The more cores, the better the card can handle graphics and computing in general.
The number of tensor cores that dynamically optimize calculations and are great at handling workloads typical of working with AI, matrix multiplication for training neural networks and data analysis.
The number of RT (Ray Tracing) cores that provide high rendering accuracy.
By the way, NVIDIA does not always indicate the exact number of CUDA, RT and tensor cores. For comparison, we used data from third-party sources.
Memory size.
Memory bandwidth. These two points logically affect the performance of the video card.
VDI virtual GPU support. This point is important because VDI is often used by our customers.
Energy consumption. This is rather a fad for us: for a data center, this indicator is important when choosing a case, power for the server and rack.

Here’s what happened in numbers:

GPU	RTX A2000	RTX A4000	RTX A5000	RTX A6000	A10	A16	A30	A40	A100 PCIe
CUDA core	3328	6144	8192	10752	9216*	1280×4	3804*	10752	6912*
Tensor cores	104	192	256	336	288*	40×4	224*	336	432*
RT core	26	48	64	84	72	10×4	—	84	—
Memory size (GB)	6 GDDR6	16 GDDR6	24 GDDR6	48 GDDR6	24 GDDR6	16×4 GDDR6	24 HBM2	48 GDDR6 ECC	40 HBM2
Memory bandwidth (Gb/s)	288	448	768	768	600	200×4	933	696	1555
VDI vGPU support	—	—	+	+	+	+	+	+	+
Maximum power consumption (W)	70	140	230	300	150	250	165	300	250

_{Data not provided by NVIDIA, taken from open third-party sources (pny. eu, techpowerup.com).}

What can you conclude from this plate

For the RTX Ax000 line, the performance grows almost linearly with the model index.

A16 is four graphics cards in one. NVIDIA is positioning the device as a dedicated VDI solution.

A30, at first glance, less productive than A10, but the HBM2 memory type has a higher bandwidth. NVIDIA is positioning the A30 as an AI solution. For both devices, the company does not publish data on the number of tensor and other cores (characteristics obtained from third-party sources).

Compared to other video cards in the table, the top PCIe A100 solution has the maximum memory bandwidth and the maximum number of tensor cores, which is expected. Obviously, the main purpose of this GPU is to work with artificial intelligence and complex calculations. This is the highest performing graphics card in NVIDIA’s lineup to date, especially the 80GB version in the SXM form factor. But the latter is soldered on the board, and for reasons of unification, we considered only the option in the PCIe form factor.

NVIDIA A4000.

How much is

Following the already voiced formula for the choice of components, consider the prices. It is difficult to write about them in 2021, which was remembered for the chip crisis and constant supply disruptions.

There will be no exact figures for two reasons. First, it is a trade secret. Secondly, and most importantly, since the cards were announced in the spring, prices have changed (and I’m sure they will continue to change in the future).

Let’s use the following approach: let’s take the A5000 GPU as a standard — its price in the comparative table will be equal to 1 «parrot». I will present the prices for the rest of the cards through the ratio to the price of the A5000. A10 and A16 are in a close price range, so they «cost» the same.

GPU	A2000	A4000	A5000	A6000	A10	A16	A30	A40	A100
Price	0. 2	0.5	1	2	1	1	1.6	1.7	3.3

At this stage, the ratio of prices and declared characteristics is expected. The first candidate to be added to the line of Selectel video cards, for the role of the junior model, is the A2000. Also of interest is the parity between the A5000, A10 and A16.

The underside of our GPUs.

Let’s move on to testing the performance of the applicants.

Video card testing

Conducting equipment tests is a common practice for Selectel. We use a large amount of hardware in various company products, so we test it both for compatibility with each other and software, and for performance.

For this we have our own «laboratory» — Selectel Lab. We even provide some equipment to customers for free testing in their projects. From fresh examples: we are giving a real monster DGX A100 with 8 video cards of the same name for testing. You can read more about its benchmark at the link.

To test the new graphics cards, we have assembled test servers with two powerful processors from Intel and enough RAM.

Here is one of the servers to test.

The specifications are as follows:

2 × Intel® Xeon® Gold 6240: 18 cores @ 2.6 GHz
192-384 GB DDR4;
240-480 GB SSD SATA;
1 × selected GPU

The benchmarks we have chosen are:

GeekBench 5 is a general benchmark that simulates task execution and measures GPU performance.

AI-benchmark is a performance test that measures the speed of learning and applying various neural networks to recognition and classification tasks.

V-Ray Benchmark is a test to check rendering speed.

ffmpeg NVENC is a video transcoding performance test.

Test results are presented in the table. We identified leaders for each item.

	Model	RTX A2000	RTX A4000	RTX A5000	A10	A30	A40	A100
GeekBench 5	OpenCL Compute Score	81 638	137 850	182 930	167 215	122 106	N/A	170 137
GeekBench 5	CUDA Compute Score	87 283	144 283	197 025	172 765	134 492	221 139	213 899
AI benchmark	Inference Score	8611	13 707	18 947	15 860	18 016	18 489	25 177
	Training Score	9 127	14 123	19 183	16 279	19385	19 265	23 775
	AI-Score	17 738	27 830	38 130	32 139	37 401	37 754	48 952
V-Ray	V-Ray Benchmark, vpaths	721	1 317	1742	1 193	897	1738	1539
ffmpeg NVENC benchmark	fps	172	173	175	N/A	N/A	157	N/A
ffmpeg NVENC benchmark	Time, s	110. 98	110.38	108.81	N/A	N/A	121.85	N/A

At the time of writing, we did not have A16 and RTX A6000 video cards on hand, so they were not included in the table. Their benchmark is planned later.

Benchmark leaders

According to the test results, the A5000 wins in terms of price-quality ratio. The best result in OpenCL Compute Score is slightly inferior to the more expensive A40 and A100 in CUDA Compute Score and is suitable for working with graphics. Second place in AI-benchmark after A100. Leader in the V-Ray rendering speed test, leader in the transcoding test. Supports VDI. Unconditionally our option, if compared with the price table.

A2000 is five times cheaper than the A5000, while still showing acceptable benchmark results for the base model. Does not support VDI, but is suitable for graphics and AI tasks.

The A4000 is mid-range between the A2000 and A5000 in terms of performance, does not support VDI, but otherwise withstands criticism in terms of price and benchmark results.

A100, as I already wrote, is the undisputed leader for working with artificial intelligence, model training, inference, data analysis and complex calculations. Ideal for Remote Desktop Infrastructure.

Other GPUs showed less results when comparing benchmarks and prices.

Final

On the podium of the winners (which, by the way, can already be ordered on the site) are four video cards. Our formula matches the RTX A2000, RTX A4000, RTX A5000 and A100.

We want to give customers the freedom to choose from low-cost single GPU servers to multi-GPU clusters. If you need a «strong» server for rendering, add A2000 to it — it will do the job for five and will not «eat» the budget. And for ambitious tasks with complex calculations, AI, large VDI projects, there is a server with eight A100s.