Nvidia GeForce RTX 5080 review: Betting the future on ‘fake frames’

“Oh, so that’s why the GeForce RTX 5080 costs less than everyone expected.”

That thought struck my mind the instant I saw where Nvidia’s new $999 graphics card fell in our gaming benchmarks. The extravagant $1,999 GeForce RTX 5090 managed to use a tantalizing mix of brute force and DLSS 4 innovation to bully its way to the top of the performance charts. The still expensive, yet more attainable RTX 5080 takes a more reserved approach, and delivers disappointing performance gains as a result. It’s barely beats the 4080 Super it’s replacing, much less the still-ferocious RTX 4090.

That makes DLSS 4, the new generation of Nvidia tech that can insert up to three AI-generated frames between every traditionally rendered frame, even more crucial to the RTX 5080. It’s a truly magical feature to play around with, sending frame rates and visual smoothness soaring, but is it enough to make up for a GPU that, frankly, delivers a poor generational performance uplift?

Yes, believe it or not — DLSS 4’s new capabilities supercharge how your games feel, imbuing even janky performers with shocking speed and snappiness.

Woof. This one’s going to be complicated. Watch Adam and Will’s video below for a benchmark-by-benchmark analysis of all the tests we’ve run. Here, we’ll focus on the key details that would-be RTX 5080 buyers need to know before dropping a cool grand on Nvidia’s latest — and sure to be controversial — enthusiast graphics card.

Nvidia’s GeForce RTX 5080 isn’t much faster than the 4080 Super

When we analyzed the RTX 50-series’ tech specs after their reveal, I pointed out that the RTX 5080 only has about 10 percent more CUDA cores than the 4080. Architecture improvements, a higher power draw, and the move to cutting-edge GDDR7 memory could also help increase performance, but the RTX 5080 wasn’t likely to be a humongous leap forward.

Unfortunately, it only provided an awkward foot-shuffle forward in our gaming benchmarks.

Across our suite — which uses a mix of different game engines, genres, and ray tracing levels — the GeForce RTX 5080 ends up just 15 percent faster at 4K resolution than the $999 RTX 4080 Super. (The vanilla 4080 launched at $1,200 before flopping and being replaced by the cheaper Super.) That’s deeply disappointing. You hope to see a 25 to 30 percent performance improvement in a new graphics card generation.

Some games perform better or worse. In Assassin’s Creed Valhalla the uplift was only 10 percent. In Cyberpunk 2077, a game developed with deep Nvidia collaboration, the RTX 5080 ran 32 percent faster than the 4080 Super. But in general, expect to see about a 15 percent uplift in most games.

The GeForce RTX 5080’s raw performance is even more disappointing if you’re looking to pair it with a high refresh-rate 1440p monitor. At that resolution, the 5080 is just 11.5 percent faster than the 4080 Super on average.

From a raw rendering perspective, the GeForce RTX 5080 is one of the worst generational upgrades in recent memory. This graphics card is barely faster than its predecessor, and the RTX 5080 falls well behind last generation’s RTX 4090. That aging behemoth runs 15 percent faster than the 5080 at 4K.

Sigh. And because of that…

DLSS 4 will make or break the RTX 5080

Watch the Full Nerd gang discuss their DLSS 4 and Multi Frame Gen experience in the time-stamped video above.

The GeForce RTX 5080 has an ace up its sleeve, though: Nvidia’s flat-out awesome new DLSS 4 technology.

More specifically, the new Multi Frame Gen feature exclusive to GeForce RTX 50-series graphics cards. It builds atop the Frame Generation feature introduced in the 40-series. While the older Frame Generation inserts a single AI-generated frame between two “traditionally” rendered frames to increase frame rate, Multi Frame Gen inserts up to three AI generated images between frames to send frame rates soaring even higher. Meanwhile, Nvidia’s mandatory-for-Frame-Gen Reflex technology helps to drive down the latency introduced by the AI frames.

It feels wonderful. Enabling MFG makes supported games (like Cyberpunk and Alan Wake 2) look and feel so much smoother. Consider this testimonial from our 5090 review:

“PCWorld contributor Will Smith, who is working on a deeper dive into DLSS 4, delivers even stronger praise: He reports that turning on DLSS 4 makes Star Wars Outlaws, a fun third-person game prone to performance concerns, feel just as good as the legendary Doom 2016, which many gamers consider the paragon of fast-action shooters. “It’s like a whole new game,” he said.”

I’ve spent time tooling around the streets of Night City with MFG active in Cyberpunk 2077. It’s astonishing how much more smooth and fast everything feels, especially in a game that runs damned smooth and fast to begin with. It’s revolutionary.

Multi Frame Gen isn’t free, however, as the excellent analysis by Hardware Unboxed above drives home. Your inputs don’t affect the AI frames, only the traditionally rendered ones. Nvidia Reflex does an admirable job of keeping latency — responsiveness — around native levels even when churning out maximum AI frames. The full 4x MFG mode only adds a handful of milliseconds of latency compared to native rendering, in exchange for DLSS 4’s delightful visual smoothness. But the way the game feels still relies on those traditionally rendered frames.

That insight unlocks several others. But the key one is this: You need a high base frame rate — 60fps or more, ideally 80fps or more — before turning on MFG to keep your games feeling “right.” If you go too much below that, the input lag becomes much more noticeable.

Fortunately, while the RTX 5080 offers only a mediocre upgrade over the 4080 Super in raw performance, the performance on offer is still more than enough to drive that 60- to 80fps “base rate” that’s so important for reasonable latency with Multi Frame Gen. Yes, Cyberpunk 2077 running at 60fps but augmented to run at 240fps with MFG/Reflex still feels like it’s running around 60fps, but constant frame pacing that comes with running a game so fast just looks and feels gooooood, man.

Frame Generation isn’t for everyone. That said, I am sensitive to latency. I am a monitor collector who has a 360Hz 1080p panel solely to play competitive shooters. I am a freak who can tell that a game running at 240fps with MFG doesn’t feel like a game running at native 240fps, and because of that I’d never use Frame Generation in multiplayer games. But the tight frame pacing and smoother visuals MFG provides makes gorgeous single-player games look and feel so much better that I unabashedly recommend using it when available, even if it feels minimally “off” at first. It’s that good.

Better yet, DLSS 4 and Multi Frame Gen will be available in 75 games and apps when the RTX 50-series launches on January 30. Some of them will be actual game updates, while you’ll need to force DLSS 4 via the Nvidia app in others. But you’ll have plenty of games to play around in.

That’s great, because DLSS 4 is the defining feature of the RTX 5080. If you’re not interested in AI frames whatsoever, the old 4080 Super would probably be a better purchase for you, weirdly enough.

Adam Patrick Murray / Foundry

The RTX 5080 is no RTX 4090 rival

…well, unless you’re running a DLSS 4 game with Multi Frame Generation active.

But other than that, the RTX 4090 still demolishes the RTX 5080. It winds up being 15 percent faster than the RTX 5080 on average at 4K resolution. Usually, the xx80 card of a generation clearly beats the xx80 Ti or xx90 tier from the previous generation, but that’s very much not the case here.

The newer card’s 16GB of GDDR7 memory doesn’t hold a candle to the RTX 4090’s massive 24GB memory buffer either. The 4090’s mondo capacity makes it beloved by content creators and AI pros alike. If you can get by with 16GB, the RTX 5080’s next-gen Blackwell GPU architecture managed to come close to the RTX 4090 in our Procyon AI text generation and Adobe Premiere Pro tests, for significantly less money. But 16GB just isn’t enough for many pros working on the most strenuous tasks these days — and while our gaming testing didn’t stress the capacity, 16GB seems awfully skimpy for a $1,000 graphics card in 2025.

Damn, Nvidia’s RTX 5080 Founders Edition is nice

If you’re lucky enough to snag an RTX 5080 Founders Edition, you won’t be disappointed. Like the RTX 5090 Founders Edition, the RTX 5080 FE wields delightfully advanced engineering tricks to squeeze the GPU’s might into a svelte true two-slot package. That means you can cram it into tiny small form-factor builds. Drooooool.

Watch GeForce’s product manager explain its intricacies in our video from CES 2025 below.

You’ll no doubt find a slew of monstrous, triple-slot custom RTX 5080s from Nvidia’s usual board partners. Their expansive heft and cooling setups will surely keep the 5080 frostier and quieter than Nvidia’s space-constrained Founders Edition, but expect to pay considerably more in return.

Given its small performance upgrade, I’m not sure that paying a steep premium for a custom RTX 5080 makes much sense unless you’re a fanatic about sound and temperatures. The Founders Edition isn’t especially hot or cranky sounding, anyways.

More power, Captain!

The RTX 5080 has a higher TGP rating than the 4080, which means it draws more power. That extra power can be used to help juice clock speeds, but some of the increase comes from the debut of cutting-edge GDDR7 VRAM fast enough to melt faces.

It’s a nothingburger in reality. The RTX 5080 consumed about 8 percent more energy than the 4080 Super in our test that logs power throughout a benchmark run. But that slightly higher power draw does pack a tangible difference: Nvidia says an 850-watt power supply is required to run the RTX 5080, a 100W increase over the 4080.

Should you buy the GeForce RTX 5080?

Only consider the RTX 5080 if you buy into Nvidia’s AI-fueled vision of the future.

Adam Patrick Murray / Foundry

DLSS 4’s Multi Frame Generation feature must be seen (and felt) to be believed. On PCWorld’s Full Nerd podcast, we compared the leap from Single Frame to Multi Frame Generation to the leap from DLSS 1 to DLSS 2. When both technologies first came out, they showed promise but had plenty of rough edges. With DLSS 2, gamers agreed that Nvidia nailed it. And while it’s not quite perfect, Multi Frame Generation nails it. Once more gamers get their Dorito-stained paws on RTX 50-series cards, and are able to tool around with MFG in 75+ games and apps, I wouldn’t be surprised if all the furor over “fake frames” online dies down quite a bit. It’s a literal game changer.

But Nvidia is in trouble this generation if the masses don’t embrace Multi Frame Generation. Because when it comes to traditional gaming performance, the RTX 5080 is no game changer.

It’s a pretty damned terrible generational upgrade, actually. Eking out a mere 11 to 15 more render performance than the RTX 4080 Super, at the same price, at a higher power draw, isn’t compelling whatsoever. It can’t come anywhere close to last gen’s 4090. If you don’t like AI-generated frames — maybe you’re sensitive to latency, or you focus on competitive games, or you loathe the idea of AI frames potentially introducing visual glitches — I’d even go so far as to suggest picking up a 4080 Super to get roughly comparable performance for less cash.

Remember: The RTX 3080 beat the RTX 2080 by 60 to 80 percent when it launched earlier this decade, and it did so for just $700. Then Nvidia jacked the price of the vanilla RTX 4080 by $500 dollars for a 30 percent performance increase, leading to poor sales rectified only by the launch of the 4080 Super at $999. With the RTX 5080 barely outpacing that, the RTX 5080 would have been immensely more compelling at a couple hundred dollars cheaper. Two generations after the RTX 3080, Nvidia has truly devastated the xx80 tier’s value in recent memory. Upgrading from the 3080 to a 5080 will only get you about 40 to 45 percent more performance, for a price tag that’s 42 percent higher. That’s not progress.

If Nvidia didn’t have MFG in tow, this would’ve been a scathing review for the RTX 5080 itself. But boyyyyy does DLSS 4’s new tricks feel great. Multi Frame Generation makes Star Wars Outlaws, a notoriously janky game, feel just as good as Doom 2016. Cyberpunk’s neon Night City feels so much more alive when you’re racing around at a buttery-smooth 240Hz+, or over 150fps even with the game’s nuclear RT Overdrive Mode active.

And that’s the promise Nvidia needs gamers to buy into for the GeForce RTX 5080 — heck, perhaps this entire RTX 50-series generation. Are you willing to embrace “fake frames” and dip your toes into experiences that aren’t currently possible with traditional rendering alone? If so, this GPU provides enough grunt to fuel those adventures in 4K and 1440p alike.

If not, the RTX 5080 is one of the most disappointing GPU releases in a long time. It’s probably best to save your cash.

Me? I’m into the vision. But I wish Nvidia imbued the RTX 5080 with more raw rendering firepower, so it could be a decent upgrade even for “fake frame” haters. Nvidia didn’t, alas — so now the RTX 5080’s future hangs in the balance of those 75 DLSS 4 games working correctly at launch.

If DLSS 4 and Multi Frame Generation perform like a champ when that wider availability hits, it could usher in a new era of smooth, AI-supercharged performance. But if DLSS 4 winds up plagued by visual artifacts or other issues once the floodgates open, it could instead set off an explosion of “fake frames” memes and sign a death warrant for the otherwise ho-hum RTX 5080 — perhaps even the rest of Nvidia’s 50-series lineup.

The GeForce RTX 5090 can stand alone on its own merits, but the RTX 5080 is all-in on DLSS 4. All that’s left us to see is where the chips fall.