The best graphics cards are the beating heart of any gaming PC, and everything else comes second. Without a powerful GPU pushing pixels, even the fastest CPU won’t manage much. While no one graphics card will be right for everyone, we’ll provide options for every budget and mindset below. Whether you’re after the fastest graphics card, the best value, or the best card at a given price, we’ve got you covered.
Where our GPU benchmarks hierarchy ranks all of the cards based purely on performance, our list of the best graphics cards looks at the whole package. Current GPU pricing, performance, features, efficiency, and availability are all important, though the weighting becomes more subjective. Factoring in all of those aspects, these are the best graphics cards that are currently available.
There are no new additions to our list this month, as the only new offering is the rather lackluster RTX 4060 Ti 16GB. There may be users that really need more VRAM and don’t care about the lack of compute or memory bandwidth, but gamers will generally be better served with other options. All current generation GPUs are now selling at or below their respective MSRPs, but generational pricing has trended upward in many cases.
Looking forward, we expect AMD to finish out its RX 7000-series lineup in the next month with at least the RX 7800 XT and RX 7700 — and possibly an RX 7800 non-XT as well. Nvidia has pretty much done everything we’d expect from the RTX 40-series, though there’s still room for an RTX 4050 at the bottom of the stack. Will it also offer 8GB of memory on a 128-bit interface, or could Nvidia try foisting a 96-bit 6GB card on the market? Hopefully the former rather than the latter.
Intel’s Arc Alchemist GPUs meanwhile still feel more like previous generation hardware, as they’re manufactured on TSMC N6 and compete more directly against the RTX 3060 and RX 6700 10GB instead of newer parts. However, Arc A750 priced at $199 remains a very competitive option (if you don’t mind the potential driver snafus and higher power use).
Note: We’re showing current online prices alongside the official launch MSRPs in the above table, with the GPUs sorted by performance.
Our list now consists mostly of current generation cards, with only a few previous generation parts still hanging around. Most of the previous generation parts are near the bottom of the performance list, though many of those like the Arc A750 and RX 6650 XT still offer an incredible value if you’re not after maximum performance.
The performance ranking (above) incorporates 15 games from our updated test suite, with both rasterization and ray tracing performance included. While we previously had DXR (DirectX Raytracing) as a separate column, we feel there are enough RT-enabled games now to aggregate the scores. Note that we are not including upscaling results in the table, which would skew things more in favor of Nvidia GPUs, but the DXR games at least partially account for that.
The above table is sorted by performance, which is why the RTX 4090 sits at the top, and why the RTX 4070 Ti edges past the 7900 XT. Our subjective rankings below factor in price, power, and features colored by our own opinions. Others may offer a slightly different take, but all of the cards on this list are worthy of your consideration.
Best Graphics Cards for Gaming 2023
Why you can trust Tom’s Hardware
Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
For some, the best graphics card is the fastest card, pricing be damned. Nvidia’s GeForce RTX 4090 caters to precisely this category of user. It’s also the debut of Nvidia’s brand-new Ada Lovelace architecture, and as such will represent the most potent card Nvidia has to offer… at least until the inevitable RTX 4090 Ti shows up.
If you were disappointed that the RTX 3090 Ti was only moderately faster (~30%) than an RTX 3080 in most workloads, RTX 4090 has something more to offer. Across our suite of gaming benchmarks, it’s 60% faster than the 3090 Ti on average. AMD’s RX 7900 XTX can’t touch it either, as the 4090 is 50% faster overall — though it also costs about 60% more.
Let’s be clear about something: You really need a high refresh rate 4K monitor to get the most out of the RTX 4090. At 1440p its advantage over a 3090 Ti shrinks to 49%, and it’s only 38% at 1080p — and that includes some demanding DXR games. The lead over the RX 7900 XTX also falls to only 31% at 1080p.
It’s not just gaming performance, either. In professional content creation workloads like Blender, Octane, and V-Ray, the RTX 4090 is about 80% faster than the RTX 3090 Ti. With Blender, it’s over three times faster than the RX 7900 XTX. And don’t even get us started on artificial intelligence tasks. In Stable Diffusion testing, besides being more difficult to get things running on AMD GPUs, the RTX 4090 was about seven times faster than the RX 6950 XT. There are numerous other AI workloads that currently only run on Nvidia GPUs. In other words, Nvidia knows a thing or two about professional applications, and the only potential problem is that it locks improved performance in some apps (like many of those in SPECviewperf) to its true professional cards, i.e. the RTX 6000 48GB.
AMD’s RDNA 3 response to Ada Lovelace might be a better value, at least if you’re only looking at rasterization games, but for raw performance the RTX 4090 reigns as the current champion. You might also want a CPU and power supply upgrade to get the most out of the 4090. At least we’re routinely able to find RTX 4090 cards selling for close to the $1,600 launch MSRP these days.
Read: Nvidia GeForce RTX 4090 Review
The Red Team King is dead; long live the Red Team King! AMD’s Radeon RX 7900 XTX has supplanted the previous generation RX 6950 XT at the top of the charts, with a price bump to match. Ostensibly priced at $999, it sold out almost immediately, but supply has now caught up to demand. And there’s good reason for the demand, as the 7900 XTX comes packing AMD’s latest RDNA 3 architecture.
That gives the 7900 XTX a lot more potential compute, and you get 33% more memory and bandwidth as well. Compared to the 6950 XT, on average the new GPU is 40% faster at 4K, though that shrinks to 30% at 1440p and just 24% at 1080p. It also delivers that performance boost without dramatically increasing power use or graphics card size.
AMD remains a potent solution for anyone that doesn’t care much about ray tracing — and when you see the massive hit to performance for often relatively mild gains in image fidelity, we can understand why many feel that way. Still, the number of games with RT support continues to grow, and many of those also support Nvidia’s DLSS technology, something AMD hasn’t fully countered even if FSR2 can at times come close. If you want the best DXR/RT experience right now, Nvidia still wins hands down.
AMD’s GPUs can also be used for professional tasks, but here things get a bit hit and miss. Certain apps in the SPECviewperf suite run great on AMD hardware, others come up short. However, if you want to do AI or deep learning research, there’s no question Nvidia’s cards are a far better pick. For this generation, the RX 7900 XTX is AMD’s fastest option, and it definitely packs a punch. If you’re willing to step down to the 7900 XT, that’s also worth considering (see below), as it tends to be priced better right now.
Further Reading:
AMD Radeon RX 7900 XTX review
The Radeon RX 6600 takes everything good about the 6650 XT below, then scales it back slightly. It’s about 20% slower overall, and factoring in DXR games it just barely edges past the RTX 3050. Discount DXR — as you should with this level of hardware — and it ends up being more like 25% faster than the 3050 and just slightly behind the RTX 3060 and Arc A750. Of course it’s also slower than the new RTX 4060, but since it costs over $100 less, we can let that slide.
Prices have dropped as low as $179 now, making this an extremely attractive option if you’re on a budget. AMD’s $329 MSRP was rather high at launch, but with Ethereum mining dead, there’s a glut of midrange GPUs available. If you’re not interested in the new round of GPUs from AMD and Nvidia and just want a decent mainstream solution, this is a great option.
Midrange graphics cards are a competitive arena, and the RX 6600 goes up against both the (now discounted) RTX 3050 as well as previous generation RTX 20-series GPUs. It ends up delivering near-RTX 2070 performance in our testing, at least in non-ray tracing scenarios. With ray tracing enabled, however, it struggles badly, barely averaging 30 fps in our DXR test suite at 1080p medium.
If you’re not worried about ray tracing, the RX 6600 definitely warrants a look. AMD’s Infinity Cache does wonders for what otherwise looks like a somewhat underpowered GPU, and the card only needs about 130W, far less than most competing GPUs. Considering the new RX 7600 costs about $80 more, that’s about a 45% increase in price for 25–30 percent more performance.
Read: AMD Radeon RX 6600 Review
AMD’s RX 6650 XT is a refresh and replacement for the RX 6600 XT, offering slightly more performance at basically the same price. It’s the second best value overall, right behind the RX 6600, but gives a bit more oomph for your fps. It’s also nearly as fast as the new RX 7600, at least in current testing, trailing by just 3% — future game and driver updates could increase the gap, however.
Performance also ends up slightly above the previous gen RX 5700 XT, which is impressive considering the memory bus has been cut in half to just 128 bits. There’s understandable concern with the 8GB of VRAM, however, but AMD and Nvidia are both sticking with 8GB on their ~$300 cards for the latest generation. If you encounter a game that runs out of VRAM, we recommend dropping texture quality a notch or two. You probably won’t even notice the difference at 1080p!
As noted on the other AMD GPUs, ray tracing performance isn’t a strong selling point. Several games that we tested with DXR support couldn’t even do 30 fps at 1080p ultra, though most were playable at 1080p medium (at which point, you might need to rethink enabling DXR). Nvidia’s RTX 3060 for example is about 35% faster (more in games like Cyberpunk 2077 and Minecraft), without using DLSS. With DLSS Quality mode, Nvidia would get another 30–40 percent faster. FSR2 doesn’t really fix that either, since it provides a similar boost in performance to all GPUs, plus it doesn’t look as good.
Overall, the launch price and reception of the RX 6650 XT wasn’t great, just like the new RX 7600. That’s reflected in current online pricing, which has helped turn things around quite a lot. The RX 6650 XT has an official $399 MSRP, but it’s now available starting at $230. Against the RTX 3050, which carries a similar online price, the RX 6650 XT looks awesome.
Further Reading:
AMD Radeon RX 6650 XT Review
We’re wrapping up the transition between old and new generation GPUs, and it leads to some difficult decisions. The RTX 4070 didn’t blow us away with extreme performance or value… but it’s generally equal to the previous generation RTX 3080, comes with the latest Ada Lovelace architecture and features, and costs $100 less (though the RTX 3080 and above are effectively discontinued now).
Nvidia’s not going to win any awards for offering a great value, as it charges the absolute maximum it feels it can get away with. At the same time, it looks better than a lot of other possibilities. In our overall performance rankings, it’s basically tied with AMD’s RX 6950 XT — slower in rasterization, faster in ray tracing, plus it has DLSS support — but the real advantage is in power requirements. With a 200W TGP, it uses about half as much electricity as many factory overclocked RX 6950 XT cards.
Nvidia is keen to point out how much faster the RTX 40-series is, once you enable DLSS 3 Frame Generation. As we’ve said before, these generated frames aren’t the same as “real” frames and increase input latency. It’s not that DLSS 3 is bad, but we prefer to compare non-enhanced performance, and in terms of feel we’d say DLSS 3 improves the experience over the baseline by 10–20 percent, not the 60–100 percent you’ll see in Nvidia’s performance charts.
The RTX 4070 has been pretty much available at MSRP since it launched, which again speaks to the lack of demand for “upper mainstream” parts that carry high-end pricing. Factory overclocked cards with extra RGB cost the usual $20–$50 extra. We still can’t help but feel the cards are a bit overpriced, but there’s no question the RTX 4070 easily beats the previous generation RTX 3070 Ti in performance at the same $599 price point — it’s about 23% faster overall.
Further Reading:
Nvidia GeForce RTX 4070 Review
With prices heading up on many previous generation cards, we’re left looking for the best options. AMD’s new RX 7900 XT generally beats the RTX 4070 Ti (below) in rasterization performance but trails by quite a bit in ray tracing games — though thankfully the extra $100 in MSRP has now effectively evaporated. That brings price parity from AMD, and you get all the latest AMD RDNA 3 architectural updates.
AMD also doesn’t skimp on VRAM, providing you with 20GB. That’s 67% more than the competing 4070 Ti. However, you won’t get DLSS support, and FSR2 works on Nvidia as well as AMD, so it’s not really an advantage (plus DLSS still looks better).
Something else to consider is that while it’s possible to run AI workloads on AMD’s GPUs, performance can at times be substantially slower. That’s because the “AI Accelerators” in RDNA 3 share the same execution pipelines as the GPU shaders, and FP16 or INT8 throughput is only double the FP32 rate. That’s enough for AI inference, mostly, but it only matches a modest GPU like the RTX 3060 in pure AI number crunching. Most AI projects are also heavily invested in Nvidia’s ecosystem, which makes them easier to get running.
AMD made a lot of noise about its new innovative GPU chiplet architecture, and it could certainly prove to be a game changer… in future iterations. For now, GPU chiplets are more about saving cost than improving performance. Consider that the die sizes of AD104 and Navi 31’s GPC are similar, but AMD also has to add five MCDs and you can see why it was supposed to be the more expensive card. And yet, performance still slightly favors Nvidia’s 4070 Ti overall — and that’s before accounting for DLSS and DLSS 3.
You might still consider a previous generation RX 6950 XT, RX 6800 XT, or RX 6800 if you’re after a less expensive but still fast AMD GPU. Prices dropped on many of those in response to the RTX 4070 launch, but availability and pricing can fluctuate a lot, and supply may dry up soon. If you’re going to spend $500 or more, we’d rather pay a bit more for the latest generation hardware and features (meaning, this or an RTX 4070), rather than picking up something that’s now more than two years old.
Read: AMD Radeon RX 7900 XT Review
The RTX 4070 Ti is the higher performance, higher price Nvidia AD104 card, and by comparison it makes the above RTX 4070 look like a great idea. The 4070 Ti drew a lot of flak for its high price at launch, as it’s $200 more expensive than the outgoing RTX 3070 Ti. While it’s definitely the faster card of the two, it’s also not that much better than the previous generation RTX 3080. You’re now given the option to pay more for more performance and the latest Ada Lovelace architecture and features.
This is basically half of an RTX 4090, for half the price. The previous generation 3070 Ti was more like 65% of the RTX 3090, at 40% of the price. Ah, how we miss the good old days… except cryptocurrency mining totally messed up retail prices during the late-2020 to mid-2022 time frame, so today’s “not great” prices are actually much better than the garbage we were dealing with a year or two ago.
It’s not just a question of gaming performance, however. Nvidia’s GPUs are the de facto choice for anyone doing AI research, at least as far as consumer hardware is concerned. If that’s something you’re interested in, don’t even bother with AMD or Intel GPUs, at least not right now. Most GitHub AI projects are built for Nvidia GPUs, and AMD or Intel forks only show up much later, if at all. That’s likely part of why Nvidia feels it can charge a premium.
The 4070 Ti has been pretty much available at MSRP since it launched. We’re even seeing a few sales where some cards dip below the $800 MSRP — briefly, and not by much. But the above RTX 4070 has now stolen a lot of the reasons for considering the RTX 4070 Ti. Yes, the Ti is about 20% faster, but it costs 33% more and still only has 12GB VRAM.
Further Reading:
Nvidia GeForce RTX 4070 Ti Review
AMD’s previous generation midrange/high-end offering follows the usual path. Take the top Navi 21 GPU and then cut down the various functional units to create a smaller die that can sell at lower prices. That’s AMD’s Navi 22 and the RX 6700 XT. The RX 6750 XT is the same GPU, with slightly higher clock speeds, memory speeds, and power consumption — about 5% faster overall. Give some thought to the RX 6700 10GB as well, which offers less performance but also costs less. We still don’t have an RX 7700-class replacement, either, though given what we’ve seen from the RX 7600 we’re not sure how much better such a card will eventually be.
The 6700 XT has the same number of GPU cores as the previous generation RX 5700 XT, but significantly higher clock speeds and more cache give it about a 30% boost to performance (at higher settings and resolutions, at least). During testing, the RX 6700 XT hit clock speeds in excess of 2.6GHz while gaming — and that’s at stock, on the reference card. Factory overclocked models can push that closer to 2.7GHz, still without cooking the GPU.
In our performance testing, the RX 6700 XT trades blows with the RTX 3060 Ti. It’s a bit faster in rasterization performance, but substantially slower in ray tracing games. That’s despite having 50% more VRAM, though newer games may favor the RX 6700 XT more. The good news is that the going price of around $310 lands below Nvidia’s card, which is now on clearance given the RTX 4060 Ti has arrived. (It’s about $333.) Don’t count on inventory of the 3060 Ti sticking around for long, however.
We expect a mainstream RX 7700 replacement in the July to August timeframe, but a lot of that will depend on how quickly previous generation cards get cleared out. The RX 7600 is now out, but with only 8GB of VRAM and a higher generational price. It’s not really that great, in other words, and we could see the same from an 7700-class GPU. Or not! But for now, the 6700 XT and 6750 XT are good options with 12GB VRAM.
Further Reading:
AMD Radeon RX 6700 XT review
AMD Radeon RX 6750 XT review
With the launch of the RTX 4060, Nvidia has just about wrapped up this generation of desktop graphics cards based on the Ada Lovelace architecture. There’s still potentially a desktop 4050 in the works, as well as refresh parts if needed (i.e. we could see a new variant of the RTX 20-series ‘Super’ models), but for now this is as low as Nvidia goes.
There are certainly drawbacks. Nvidia opted to cut down the memory interface to just 128 bits, which in turn limits the memory capacity options. Nvidia could do a 16GB card if it really wanted, but 8GB is the standard configuration and we don’t expect anything else — only the 4060 Ti will get the double VRAM option. It also has an x8 PCIe interface, though in practice that shouldn’t matter.
The good news is that, as promised, performance is better than the previous generation RTX 3060. There are edge cases in some games (meaning, 4K at max settings) where the 12GB on the 3060 can give it the lead, but performance is already well below the acceptable level at that point. As an example, Borderlands 3 ran at 26.5 fps on the 4060 versus 28.9 fps on the 3060 at 4K Badass settings; neither is a great experience, even though the 3060 is technically faster. At the same time, the 4060 can’t catch the RTX 3060 Ti, at least not without DLSS 3 and Frame Generation.
That’s the other benefit, of course: You get all the latest Ada features, including DLSS 3 support. Also, the power draw is just 115W for the reference model, and typically won’t exceed 125W on overclocked cards. Oh, and most RTX 4060 cards won’t bother with the questionable 16-pin power connector and adapter shenanigans.
As an alternative view, this is an upgraded RTX 3050, with the same 115W TGP and 60% better performance. Too bad it costs $50 extra, though the 3050 was mostly priced at $300 and above until the past few months.
Read: Nvidia GeForce RTX 4060 Review
AMD’s Navi 21 GPUs like the Radeon RX 6800 XT can still represent a better value than some of the latest additions, though the supply of the various models including the vanilla 6800, 6800 XT, 6900 XT, and 6950 XT seems to fluctuate a lot. There’s now only a $10–$20 gap between the 6800 XT and the 6800, making this an easy pick. But don’t be surprised if the RX 6800 drops in price again to compete, as AMD looks to clear out any remaining inventory.
All of those GPUs we just listed use AMD’s previous generation RDNA 2 architecture. It was fine when it first arrived in late 2020, and the GPUs still run most games well today. Just don’t expect exceptional ray tracing performance.
At current prices, the RX 6800 XT currently starts at around $500, putting it right between Nvidia’s new RTX 4070 and RTX 4060 Ti in pricing and matching the unexciting RTX 4060 Ti 16GB. It has 16GB of VRAM, plus a 256-bit interface and lots of L3 cache. Performance overall ends up being 10% below the 4070 and 15% above the 4060 Ti. Keep an eye on other Navi 21-based graphics cards, though, as sometimes a faster card will end up with a lower price.
A big part of AMD’s performance comes thanks to the massive 128MB Infinity Cache. It improves the effective bandwidth by 119%, according to AMD. Few if any games currently need more than 16GB, so the 6800 XT is in a great position in that area. Compute requirements in some games have gone up, however, so the 16GB doesn’t guarantee great performance on its own.
What’s not to like? The ray tracing performance is mediocre, due to AMD’s RDNA2 architecture and the lack of hardware BVH traversal (it does some of the processing for BVH on the shader cores is our understanding). There are no Tensor cores or DLSS either, though FSR2 at least partially makes up for that. With RDNA 3 cards now shipping, we also have to wonder how long it will be before something like an RX 7700-series arrives with better performance and efficiency than this two years old GPU — and hopefully at a lower price (but don’t hold your breath)!
Further Reading:
AMD Radeon RX 6800 XT and RX 6800 Review
AMD Radeon RX 6900 XT Review
AMD Radeon RX 6950 XT Review
Testing the Intel Arc A750 was a bit like dealing with Dr. Jekyll and Mr. Hyde. At times, performance looked excellent, sometimes surpassing the GeForce RTX 3060. Other times, Arc came up far short of expectations, trailing the RTX 3050. The drivers continue to improve, however, and with prices now starting at $200, this represents an excellent value — just note that some of the cost savings will ultimate show up in your electrical bill, as it’s not as efficient as the competition.
There are some compromises, like the 8GB of VRAM — the A770 Limited Edition doubles that to 16GB, but also costs over $100 extra. Intel’s A750 also has to go up against AMD’s RX 6650 XT, which is the primary competition at this price. Depending on the game, performance may end up favoring one or the other, though Intel now holds the overall edge. Ray tracing tends to favor Intel, while rasterization games are more in the AMD camp.
Intel was the first company to deliver hardware accelerated AV1 encoding and decoding support, and QuickSync continues to deliver an excellent blend of encoding performance and quality. There’s also XeSS, basically a direct competitor to Nvidia’s DLSS, except it uses Arc’s Matrix cores when present, and can even fall back into DP4a mode for non-Arc GPUs. But DLSS 2 still comes out on top, and it’s in far more games.
The Arc A750 isn’t a knockout blow, by any stretch, but it’s also nice to have a third player in the GPU arena. The A750 competes with the RTX 3060 and leaves us looking forward to Intel’s future Arc Battlemage GPUs, even if they’re probably a year out. You should also check out the Arc A770 16GB, if you’re willing to give Intel a chance, though it’s a steep upsell these days.
Further Reading:
Intel Arc A750 Limited Edition Review
Intel Arc A770 Limited Edition Review
What is this, 2016? A brand-new, $399 graphics card launching with only 8GB of memory? We thought we had left that era in the past after the RTX 3060, but Nvidia seems more intent on cost-cutting and market segmentation these days. But the RTX 4060 Ti does technically beat the previous generation RTX 3060 Ti, by 10–15 percent in our testing, though there are some exceptions.
There are plenty of reasons to waffle on this one. The larger L2 cache does mostly overcome the limited bandwidth from the 128-bit interface, but cache hit rates go down as resolution increases, meaning 1440p and especially 4K can be problematic. At least the price is the same as the outgoing RTX 3060 Ti, and you do get some new features.
If you were previously looking at the RTX 3060 Ti, and you don’t want to consider an AMD or Intel alternative, this was the least expensive Ada Lovelace / RTX 40-series GPU — but the RTX 4060 above has now taken over that title. The RTX 4060 also has fewer GPU cores and less L2 cache size, so it very much ends up as a GPU that warrants its $299 price tag.
Looking at performance, the 4060 Ti generally manages 1440p ultra at 60 fps in rasterization games, but for ray tracing you’ll want to stick with 1080p — or use DLSS. Frame Generation is heavily used in Nvidia’s marketing materials, and it can provide a significant bump to your fps. However, it’s more of a frame smoothing technique as it interpolates between two frames and doesn’t apply any new user input to the generated frame.
Besides that, you get better ray tracing hardware and AV1 encoding support. Nvidia’s new mainstream GPU is also about 35% faster than AMD’s new RX 7600 in our rasterization tests, for about 50% more money. Factor in ray tracing and it’s more like a 50% increase in performance for 50% more money. That’s okay, but then the RX 7600 isn’t an awesome new offering either.
Consider this more as a slightly higher entry point for the RTX 40-series and a competent $400 solution, but a price cut would go a long way toward sweetening the deal.
Further Reading:
Nvidia RTX 4060 Ti Review
How We Test the Best Graphics Cards
Tom’s Hardware 2023 GPU Testbed
Determining pure graphics card performance is best done by eliminating all other bottlenecks — as much as possible, at least. Our 2023 graphics card testbed consists of a Core i9-13900K CPU, MSI Z790 MEG Ace DDR5 motherboard, 32GB G.Skill DDR5-6600 CL34 memory, and a Sabrent Rocket 4 Plus-G 4TB SSD, with a be quiet! 80 Plus Titanium PSU and a Cooler Master CPU cooler.
We test across the three most common gaming resolutions, 1080p, 1440p, and 4K, using ‘medium’ and ‘ultra’ settings at 1080p and ‘ultra’ at 4K. Where possible, we use ‘reference’ cards for all of these tests, like Nvidia’s Founders Edition models and AMD’s reference designs. Most midrange and lower GPUs do not have reference models, however, and in some cases we only have factory overclocked cards for testing. We do our best to select cards that are close to the reference specs in such cases.
For each graphics card, we follow the same testing procedure. We run one pass of each benchmark to “warm up” the GPU after launching the game, then run at least two passes at each setting/resolution combination. If the two runs are basically identical (within 0.5% or less difference), we use the faster of the two runs. If there’s more than a small difference, we run the test at least twice more to determine what “normal” performance is supposed to be.
We also look at all the data and check for anomalies, so for example RTX 3070 Ti, RTX 3070, and RTX 3060 Ti all generally going to perform within a narrow range — 3070 Ti is about 5% faster than 3070, which is about 5% faster than 3060 Ti. If we see games where there are clear outliers (i.e. performance is more than 10% higher for the cards just mentioned), we’ll go back and retest whatever cards are showing the anomaly and figure out what the “correct” result should be.
Due to the length of time required for testing each GPU, updated drivers and game patches inevitably come out that can impact performance. We periodically retest a few sample cards to verify our results are still valid, and if not, we go through and retest the affected game(s) and GPU(s). We may also add games to our test suite over the coming year, if one comes out that is popular and conducive to testing — see our what makes a good game benchmark for our selection criteria.
Choosing Among the Best Graphics Cards
We’ve provided a baker’s dozen of choices for the best graphics cards, recognizing that there’s plenty of potential overlap. The latest generation GPUs consist of Nvidia’s Ada Lovelace architecture, which improves on the previous Ampere architecture. AMD’s RDNA3 architecture likewise takes over from the previous RDNA2 architecture offerings, though we have a few previous generation cards still in our list. Finally, Intel Arc Alchemist GPUs have arrived and provided some competition in the budget and midrange sectors. Conveniently, Arc Alchemist, RDNA2/3, and Ada/Ampere all support the same general features (DirectX 12 Ultimate and ray tracing), though Arc and RTX cards also have additional tensor core hardware.
We’ve listed the best graphics cards that are available right now, along with their current online prices, which we track in our GPU prices guide. With many cards now costing close to MSRP, plenty of people seem ready to upgrade, and supply also looks to be improving. At the same time, a slumping economy and rising inflation seems to have lowered demand, and the supply of new GPUs isn’t being pushed quite as hard as before.
Our advice: Don’t pay more today for yesterday’s hardware. If you want an RTX 40-series or RX 7000-series graphics card, be patient and you’ll eventually be able to buy one at close to the official MSRP. They’re only high-end and extreme offerings right now, but mainstream and budget variants will inevitably arrive.
If your main goal is gaming, you can’t forget about the CPU. Getting the best possible gaming GPU won’t help you much if your CPU is underpowered and/or out of date. So be sure to check out the Best CPUs for Gaming page, as well as our CPU Benchmark hierarchy to make sure you have the right CPU for the level of gaming you’re looking to achieve.
Our current recommendations reflect the changing GPU market, factoring in all of the above details. The GPUs are ordered using subjective rankings, taking into account performance, price, features, and efficiency, so slightly slower cards may end up higher on our list.
When buying a graphics card, consider the following:
• Resolution: The more pixels you’re pushing, the more performance you need. You don’t need a top-of-the-line GPU to game at 1080p.
• PSU: Make sure that your power supply has enough juice and the right 6-, 8- and/or 16-pin connector(s). For example, Nvidia recommends a 550-watt PSU for the RTX 3060, and you’ll need at least an 8-pin connector and possibly a 6-pin PEG connector as well. Newer RTX 40-series GPUs use 16-pin connectors, though all of them also include the necessary 8-pin to 16-pin adapters.
• Video Memory: A 4GB card is the absolute minimum right now, 6GB models are better, and 8GB or more is strongly recommended. A few games can now use 12GB of VRAM, though they’re still the exception rather than the rule.
• FreeSync or G-Sync? Either variable refresh rate technology will synchronize your GPU’s frame rate with your screen’s refresh rate. Nvidia supports G-Sync and G-Sync Compatible displays (for recommendations, see our Best Gaming Monitors list), while AMD’s FreeSync tech works with Radeon cards.
• Ray Tracing and Upscaling: The latest graphics cards support ray tracing, which can be used to enhance the visuals. DLSS provides intelligent upscaling and anti-aliasing to boost performance with similar image quality, but it’s only on Nvidia RTX cards. AMD’s FSR works on virtually any GPU and also provides upscaling and enhancement, but on a different subset of games. New to the party are DLSS 3 with Frame Generation and Intel XeSS, with yet another different subset of supported games — DLSS 3 also provides DLSS 2 support for non 40-series RTX GPUs.
Graphics Cards Performance Results
Our updated test suite of games consists of 15 titles. The data in the following charts is from testing conducted during the past several months. Only the fastest cards are tested at 1440p and 4K, but we do our best to test everything at 1080p medium and ultra.
For each resolution, the first chart shows the geometric mean (i.e. equal weighting) for all 15 games. The second chart shows performance in the nine rasterization games, and the third chart focuses in on ray tracing performance in six games. Then we’ll have the 15 individual game charts, for those who like to see all the data.
AMD’s FSR has now been out for about two years now, with FSR 2.0 now having surpassed the year mark. Nvidia’s DLSS 2 has been around since mid-2019, while Intel’s XeSS formally launched in October 2023. Twelve of the games in our test suite support DLSS 2, five more support DLSS 3, five now support FSR2, and four support XeSS. However, we’re running all of the benchmarks at native resolution for these tests. We have a separate article looking at FSR and DLSS, and the bottom line is that DLSS and XeSS improve performance with less compromise to image quality, but FSR2 works on any GPU.
The charts below contain nearly all of the current RTX 40/30-series, RX 7000/6000-series, and Intel’s Arc A-series graphics cards. Our GPU benchmarks hierarchy contains additional results for those who are interested, along with performance testing from our 2020-2021 suite running on a Core i9-9900K. The charts are color coded with AMD in red, Nvidia in blue, and Intel in gray to make it easier to see what’s going on.
The following charts are up to date as of August 22, 2023. Nearly all current generation and previous generation GPUs are included.
Best Graphics Cards — 1080p Medium
Best Graphics Cards — 1080p Ultra
Best Graphics Cards — 1440p Ultra
Best Graphics Cards — 4K Ultra
Power, Clocks, and Temperatures
Most of our discussion has focused on performance, but for those interested in power and other aspects of the GPUs, here are the appropriate charts.
Finding Discounts on the Best Graphics Cards
With the GPU shortages mostly over, you might find some particularly tasty deals on occasion. Check out the latest Newegg promo codes, Best Buy promo codes and Micro Center coupon codes.
Want to comment on our best graphics picks for gaming? Let us know what you think in the Tom’s Hardware Forums.
MORE: HDMI vs. DisplayPort: Which Is Better For Gaming?