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

Price When Reviewed

1169 Euro

Best Prices Today: Nvidia GeForce RTX 5080 Founders Edition

“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