Nvidia GeForce RTX 5080 Founders Edition

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

The standout in any given new line of products is typically the best-provisioned, fastest, and most expensive one. That’s been the case so far with Nvidia’s GeForce RTX 50-series graphics cards and the Nvidia GeForce RTX 5090. That 5090 card is a rock star by any measure, so long as you don't look too closely at the price. But most mortals bound to things like budgets should pay more attention to Nvidia's GeForce RTX 5080. This graphics card may not be the top model, but it's far more affordable, priced at $999 versus the GeForce RTX 5090’s $1,999. Its performance is also exceptional, which begs the question: Do you really need a faster graphics card than this? Our Editors' Choice award for enthusiast-grade graphics cards is our answer.

The GB203 GPU: The 'Blackwell' for Gamers

The Nvidia GeForce RTX 5080 contains an Nvidia GB203 graphics processing unit (GPU) die based on the company’s new "Blackwell" architecture. Despite the general hype for Nvidia's GeForce RTX 50 series, the RTX 5080 has been overshadowed by the Nvidia GeForce RTX 5090 until now. It would be hard not to be, given that the Nvidia GeForce RTX 5090 features a titanic graphics chip with more than 92 billion transistors, 21,760 CUDA cores, 32GB of GDDR7 memory, and a 512-bit memory interface.

The RTX 5090 pushes out the boundaries of what’s possible from a graphics chip, properly earning the most attention. However, the Nvidia GeForce RTX 5090 is also prohibitively expensive, and it is likely to be gobbled up by large companies and developers for its computing and AI performance, leaving few cards available for gamers or at least pushing up prices.

I don’t see this as much of a problem, as most folks should instead look at the GeForce RTX 5080 for a high-end, next-generation Nvidia GPU. The GB203 graphics chip is roughly half the size of a GB202 chip, and the Nvidia GeForce RTX 5080 is also akin to half an RTX 5090 in many ways. The Nvidia GeForce RTX 5080 has 10,752 CUDA cores, 16GB of GDDR7, and a 256-bit memory interface, all of which are either precisely or almost exactly half of what you’ll find on the RTX 5090.

Two aspects of the RTX 5080 are better than the RTX 5090 other than its price. First, its peak power consumption is way lower, at 360 watts versus the RTX 5090’s 575W. While it's not a primary concern in desktops for most users, a lower power draw is ideal for many reasons. This is especially true if it means you don't have to upgrade your power supply to support a new card. (The RTX 5090 really needs a 1,000W-plus model.) The second thing is the clock speeds, which are much higher on the RTX 5080 to make up, somewhat, for the reduced core count.

Naturally, as it uses a Blackwell-based graphics chip, the RTX 5080 benefits from all the same performance gains that architecture provides over the last-gen Nvidia "Ada Lovelace" architecture. The GPU will also benefit from features like Nvidia's RTX Neural Shaders once they arrive in supported games, and it can employ DLSS 4 in games that support the new frame-generation technology to generate extra frames if you need the performance boost. If you want to know more about those features, you should read our other articles linked here, where we have discussed these features directly. But for now, let's move on to the card we tested.

Design: A Look at the Nvidia GeForce RTX 5080 Founders Edition

From one perspective, this Founders Edition card directly from Nvidia feels like a super-premium, weighty product; it is the same size as the Nvidia Geforce RTX 5090. However, compared with the last-gen Nvidia GeForce RTX 4080 or the Nvidia GeForce RTX 4080 Super, it actually feels relatively small.

That the design is so similar to Nvidia’s GeForce RTX 5090 is promising. (Also see our original hands-on with the RTX 5090.) That card stayed well-cooled when we tested it, and the GeForce RTX 5080 has an identical thermal solution that allows the fans to push air through heatsinks and over cooling pipes that run to the left and right sides of the circuit board. This design, while more complex from a manufacturing standpoint (as it squeezes much of the key circuitry into the middle of the graphics card), should provide better cooling, as more air can flow through the card and its thermal hardware.

The rear I/O panel of the RTX 5080 Founders Edition is also unchanged, with one HDMI port and three DisplayPort ports. To power this card, you will either need a power supply with a native 12VHPWR connector (Nvidia's late-model power connector) or the adapter that Nvidia supplies with the card, which converts three 8-pin PCI Express power connectors to the single 12VHPWR connector on the top edge of the card. The adapters Nvidia is sending out with the RTX 50-series look and feel much higher quality than those supplied with Nvidia's RTX 30- and RTX 40-series cards, with much more flexible cables allowing for easier routing and bends that are less worrisome, given this cable's and connector's history. Also, the angled placement of the power-in port on the graphics card is an improvement, helping to support the power cable and make it less vulnerable to damage.

GeForce RTX 5080 Founders Edition Testing: Setup and Competition

We tested the Nvidia GeForce RTX 5080 Founders Edition on our newly built graphics card testbed that has an AMD Ryzen 9 9950X CPU installed on a Gigabyte X870E Aorus Master motherboard with a 360mm Cooler Master liquid cooler. The system has two 16GB sticks of DDR5 RAM operating in a dual-channel configuration with an AMD EXPO profile of 6,000MHz. For storage, we use two Crucial M.2 PCIe 4.0 SSDs, with one (a 2TB drive) dedicated to games and the other (4TB) holding Windows 11 and all other software that isn’t a game. A 1,500W Corsair power supply powers the PC.

Unlike the dominant Nvidia GeForce RTX 5090, the GeForce RTX 5080 has many rivals. For those with a big budget for their gaming PC, the Nvidia GeForce RTX 5090 is a competitor, as is the Nvidia GeForce RTX 4090, though both are prohibitively expensive otherwise. Also, while the RTX 5090 has no real competition in terms of performance, as you’ll soon see, the RTX 5080 seriously undercuts the RTX 4090 on the value front.

Eventually, the Nvidia GeForce RTX 5080 will also have to compete with Nvidia’s upcoming GeForce RTX 5070 Ti. However, as that $749 card hasn’t launched yet, the rest of the RTX 5080’s competition comes from last-gen opponents. The RTX 4080 Super, the RTX 4080, and the Nvidia GeForce RTX 4070 Ti Super all provide something in the way of competition via either raw performance or lower prices.

This is also true of the AMD competition, as cards like the AMD Radeon RX 7900 XTX are still quite capable; that flagship Radeon now retails for less than its original $999 launch price.

Because we are still updating our GPU testing regimen for 2025 and beyond, we could not test the Nvidia GeForce RTX 4080 Super on these new benchmarks in time for this review, though it's coming. However, it's important to note that generally, we saw single-digit performance gains from the RTX 4080 Super over the base RTX 4080 model compared here when we originally reviewed that card. You can look at it as a slightly lesser proxy for that card.

Synthetic Benchmarks

The test results we gathered from 3DMark delivered mixed messages about the RTX 5080’s performance. The Port Royal and Speed Way scores suggest that the RTX 5080 should handily outpace everything except the RTX 4090 and the RTX 5090. The Solar Bay and Steel Nomad scores also give the impression the RTX 5080 is faster, but its lead over the AMD cards and the last-gen Nvidia cards narrowed in these tests compared with Port Royal and Speed Way.

Time Spy Extreme, the oldest of the 3DMark tests we ran, painted a somewhat different picture, with the AMD Radeon RX 7900 XTX pulling relatively close to the RTX 5080. The 3DMark screen optimization tests also gave this impression, though we must point out that this chart doesn’t show results from a singular test. 3DMark has vendor-specific DLSS, FSR, and XeSS tests intended to run on Nvidia, AMD, and Intel graphics cards, respectively. The tests look identical, and as they test similar features for each company, we are including the results together like this, but take the numbers with a pinch of salt and compare just Nvidia to Nvidia, AMD to AMD.

Last, Unigine’s Superposition test showed that the AMD Radeon RX 7900 XTX may compete better with the RTX 5080 in games that use the OpenGL API. You probably want to look closer into this if you are running Linux, but all of the game tests we ran were DirectX-based, so we don’t have more information on this at the moment.

Content Creation Benchmarks

Gaming may be the first thing you think about surrounding a graphics-card upgrade, but cards can also be powerful tools for content creation work. Though it's clearly a very able workhorse, the GeForce RTX 5080 may not be a significant enough step up from last-gen cards to merit an upgrade. Take the PugetBench for Creators benchmark, which we run inside Adobe Premiere Pro 24. The RTX 5080 scored better than the last-gen competitors in the chart, but the Nvidia GeForce RTX 4080 and RTX 4070 Ti Super weren’t all that far behind. As those cards will likely be discounted now that the RTX 50 series is rolling out, they could be better-value options for this sort of work if you're coming from something older. (Or, they could be good enough if you already own one.)

The test scores we gathered from the Blender 4.3.0 rendering benchmark also suggest the last-gen RTX 4080 and RTX 4070 Ti Super could be better value options for content creation work if you can score one cheaper. They weren’t faster than the RTX 5080 here, but they were close enough that discounts on last-gen models could be a significant factor. The RTX 5080 performed better in the Chaos V-Ray 6 benchmark, and here, we did see an appealing performance gain in the RTX engine score. (The CUDA engine score should be largely discounted, as Nvidia informed us of a bug with that test on RTX 50-series cards.)

The RTX 5080 is certainly a very good engine overall for GPU-accelerated content creation work. Its performance gains just aren’t quite high enough to make it an appealing option for that work, in my opinion, if you are looking at the RTX 5080 as an upgrade from an equivalent RTX 40-series card. The last-gen cards are likely better options from a cost-for-performance standpoint, and if you need the most performance you can get for content creation, look to the RTX 5090 instead.

Procyon AI Benchmarks

UL’s Procyon AI text generation is one of the first reliable tests we have to gauge AI performance. Though this is likely more important for companies that will use AI hardware for specific tasks, it’s still fascinating to see how the AI hardware in these GPUs compares. (Testing AI performance in any holistic sense is difficult, particularly as on-card AI muscle addresses a wide range of tasks--training versus inferencing, for one thing. It is impossible with current tools to test a very broad set of AI local-processing scenarios in an all-encompassing way.)

This test presses the GPU on a series of text-creation inferencing tasks via four popular AI large language models (LLMs) and produces an overall score in each case. (Here, Procyon uses versions of Mistral, Microsoft's PHI, and two flavors of Meta AI's Llama.) It also reports how many tokens (or discrete units of text, in the context of text generation) the hardware can produce per second with each model. It also measures the time it takes to produce the first token, as getting as much work done as fast as possible is basically AI’s mission statement.

Like the content creation tests, the AI tests showed the RTX 5080 with only a slight lead over its last-gen counterparts. But the story is different here. The RTX 5080 may have produced only 11% more tokens per second than the RTX 4080 in the PHI 3.5 test, but AI is all about doing as much work as possible as fast as possible and producing results faster. You may not see parallel relative results in other AI tasks, but the RTX 5080 is clearly a step ahead of previous-gen equivalents in a broad selection of LLMs. The gains may also be greater in tasks that are measured in hours rather than seconds or minutes.

DLSS, FSR, and XeSS Benchmarks

The image quality and performance characteristics of DLSS, FSR, and XeSS aren’t quite the same. While similar, these technologies don’t work in quite the same way. This makes comparing the results directly imprecise, as the frames-per-second (fps) scores alone don’t take into consideration image quality differences. Due to this, we are conducting limited DLSS, FSR, and XeSS testing with one game going forward.

In the Black Myth: Wukong stand-alone benchmark, the RTX 5080 performed well against the competition. It was faster than the RTX 4090 at 1080p, with the two cards almost tied at 1440p and the RTX 4090 pulling ahead at 4K. The RTX 5080 was also faster than all other cards we tested except the RTX 5090. This situation was unchanged by the use of frame-generation technology, which inflated the numbers but kept all of the cards in the same relative positions.

Ray-Traced Gaming Benchmarks

Here, in games that support ray tracing, we see the potential of the Nvidia GeForce RTX 5080 on display. The gains vary significantly from game to game, depending on circumstances, but nowhere does the RTX 5080 disappoint.

Starting with Cyberpunk 2077, the Nvidia GeForce RTX 5080 held a noticeable lead over the preceding Nvidia GeForce RTX 4080 and a much more prominent lead over the AMD competition. The RTX 5080 lagged behind the RTX 4090 and RTX 5090 in this game, but this wasn’t consistent across the board.

F1 2024 showed a much more consistent performance gain for the RTX 5080 over the RTX 4080 than Cyberpunk. Whereas with Cyberpunk 2077, the RTX 5080 was between 7% and 16% faster than the RTX 4080, that performance gain stayed between 14% and 18% in F1 2024, depending on the resolution. Notably, the RTX 5080 surpassed the RTX 4090 at 1080p and tied with it at 1440p.

Far Cry 6 showed evidence of a bottleneck from the CPU or the game engine that caused the results to group together far more than expected. This helped the AMD competition to score a rare win over the RTX 5080 at 1080p, and the RTX 5080 likewise didn’t perform as well as the competition at 1440p. This could be related to early drivers for the RTX 5080. At 4K, the RTX 5080 still couldn’t beat the AMD Radeon RX 7900 XTX, which tied with it, but it was faster than the RTX 4080 at this resolution.

In Avatar: Frontiers of Pandora, the RTX 5080 returned to its usual position in the ranks ahead of most competitors. The RTX 5080 outpaced both competing AMD cards here and held a consistent lead over the RTX 4080, but it lagged behind the RTX 4090 and RTX 5090. Returnal again showed the RTX 5080 ahead of the RTX 4090 at 1080p, a near tie at 1440p, and the RTX 4090 ahead at 4K.

We have a chart included here for Call of Duty: Modern Warfare III, but we could not test the Nvidia GeForce RTX 5080 on this title. We tried multiple successively updated early graphics drivers, but the game crashed each time we ran it on the RTX 5080, so until this issue is corrected, we will not have performance results for the RTX 5080 on this game.

Raster-Only Gaming Benchmarks

Raster-only games do not support ray tracing, so they don’t take advantage of the ray-tracing hardware inside modern graphics cards. These games give AMD more advantage, as its ray-tracing hardware isn’t quite as adept as Intel’s or Nvidia’s. Sure enough, in Total War: Three Kingdoms, the AMD Radeon RX 7900 XTX came closer to matching the Nvidia GeForce RTX 5080 than in any title so far except the bottlenecked Far Cry 6. The RTX 5080 was still faster, but it didn’t enjoy quite as large of a lead.

Total War also showed signs of a CPU bottleneck at the 1080p resolution, which enabled the RTX 5080 to catch up to the 5090 and surpass the RTX 4090. The RTX 5090 was also bottlenecked at 1440p, but none of the other cards was. At 1440p, the RTX 5080 was still faster than the RTX 4090 and only fell behind the RTX 4090 at the 4K resolution. The RTX 5080 held a 25% advantage over the RTX 4080 at 4K, which is notable.

We likely hit yet another 1080p CPU bottleneck in Shadow of the Tomb Raider, though only the Nvidia GeForce RTX 5090 and Nvidia GeForce RTX 5080 were affected this time. Like other results, the RTX 5080 outperformed the RTX 4090 here at 1440p but fell behind at 4K, while none of the other competition was much of a match.

Power and Thermal Benchmarks

We use a Kill-A-Watt power meter to gauge the total power consumption of our graphics card test bed. Since only the graphics card changes, this helps to give us a rough idea of how demanding each card is on power draw. We test with four programs: two content creation apps, and two games.

In the content creation tests, the RTX 5080 pulled power numbers close to those of the RTX 4080 while drawing significantly less than the AMD competition and the RTX 5090. This changed in the in-game benchmarks, which set the RTX 5080 with a notably higher power draw than the last-gen RTX 4080. The RTX 5080 was also a little power-hungrier here than the Radeon RX 7900 XTX but significantly less demanding than the voracious RTX 5090.

Thermally, the RTX 5080 ran hotter than the RTX 4080 in content creation tests. Given both cards had a similar power draw, and that the RTX 5080 is supposed to have a new, improved thermal design, we would have expected to see the opposite. (That said, the RTX 4080 is indeed much thicker, which means more material for heat dissipation, all else being equal.) In the in-game benchmarks, the RTX 5080 ran hotter than the RTX 4080, but this makes more sense considering its 40W-higher board power rating.

Verdict: The Next-Gen, High-End Gaming Champion

The Nvidia GeForce RTX 5080 may always sit in the RTX 5090’s shadow in terms of raw performance, but if you are a gamer looking to build or upgrade a top-tier gaming PC, it should likely be the top option. If you’ve got the money and won’t miss it, go ahead and buy the Nvidia GeForce RTX 5090. But for that $1,999 you would spend on an RTX 5090, you could realistically drop $999 on an Nvidia GeForce RTX 5080 and build out the rest of the gaming PC with the remaining $1,000, or at least start with a tip-top CPU and some nice complementary components.

Given $1,999 is already more than most will feel comfortable to drop on a gaming PC, this option looks a lot more alluring for anyone that is remotely budget-conscious. Additionally, CPU and game-engine bottlenecks somewhat limit these gains in places, so getting the full potential out of your RTX 5090 will be tricky at times but easier with an RTX 5080. However, that card, too, encountered bottlenecks here and there at 1080p—not much you can do, though, when you are already running one of the fastest processors (the Ryzen 9 9950X) on the market. We would joke that Nvidia will need to start making its own CPUs to keep up with its graphics card advancements, if we weren’t worried Nvidia would take us seriously and do just that.

You'll find worthwhile alternatives to the RTX 5080 for competition, but the RTX 4090 is not one of those. It can be faster than the RTX 5080, but that advantage is far from consistent, with the RTX 5080 sometimes faster. Add to that the RTX 4090’s much higher price, and it's clearly just not a card that works in today’s market sandwiched between the RTX 5080 and the RTX 5090.

The Nvidia GeForce RTX 4080 and the RTX 4080 Super are more substantial alternatives. We haven’t tested the RTX 4080 Super on our new GPU testbed yet, but on average, it’s only about 5% faster than the standard RTX 4080, so we can well-enough estimate how it will perform. The gains the new RTX 5080 has over the RTX 4080 and the RTX 4080 Super aren’t consistent, varying from a few percentage points to a maximum of around 25%. Given that the RTX 5080 and RTX 4080 Super are both set with an MSRP of $999, it's evident that the RTX 5080 is the card you should buy. A discounted RTX 4080 Super could be a worthwhile alternative, but we'd want to see a substantial markdown before considering it.

That’s essentially the same spot that AMD’s Radeon RX 7900 XTX and Radeon RX 7900 XT also find themselves in. These cards were more consistently behind the GeForce RTX 5080, but with the proper discount, they could be compelling alternatives. It all comes down to the actual street pricing.

Considering all that, the Nvidia GeForce RTX 5080 Founders Edition is the graphics card we recommend if you're looking to play high-resolution, high-refresh-rate games, and plan to spend anything in the region of $1,000. For that, it merits our Editors' Choice award.

Francisco La Hoz contributed testing work to this review.