Nvidia GeForce RTX 5080 Review

The Nvidia GeForce RTX 5080 is an odd graphics card. It’s only a bit faster than its predecessor, the RTX 4080 Super, but it’s also significantly cheaper (for now), launching at $999 for the Founders Edition. This makes it not worth the upgrade for anyone who already has a 4080 or 4080 Super, but if you’ve been on the fence waiting for the new generation, you will get significantly better value out of this new graphics card – something that’s far from guaranteed these days.

But the small performance uplift over the RTX 4080, combined with the RTX 5090’s weak lead over the 4090 means that Blackwell – the GPU architecture behind these new graphics cards – is one of the least substantial graphics generations in years, at least since Turing and the RTX 2080 Ti. Nvidia is hoping to offset some of that with DLSS 4 and multi-frame generation, which should be genuinely useful to anyone with a high-refresh, high-resolution display, but even then you’re going to have to make peace with some of your frames being generated by AI.

Specs and Features

Just like the RTX 5090, the Nvidia GeForce RTX 5080 is built on the Blackwell GPU architecture, which is especially catered towards AI workloads. Still, it does feature more CUDA cores than its predecessor, with 10,752 spread across seven Graphics Processing Clusters, or GPC, compared to 9,728 cores in the 4080. Also similar to the 5090, Nvidia managed to fit more Streaming Multiprocessors, or SMs into each GPC, which means the RTX 5080 is able to have more CUDA cores without drastically increasing the size of the GPU.

Each of these SMs features four Tensor Cores and one RT Core, meaning you get 336 and 84, respectively. The Tensor Cores, specifically, have been a major focus for Nvidia over the last few years, and Team Green claims these 5th-generation cores are much more efficient, which has paved the way for DLSS 4, at least in part.

The other key behind DLSS 4 and its Multi-Frame Generation is a new AI Manager Processor, or AMP, built directly into the GPU. Essentially, the AMP directs work to different parts of the graphics card, something that was traditionally left to the CPU. But because the AMP is both located physically on the GPU, and is purpose-built for this, it’s able to schedule work much more efficiently than a traditional processor.

The Nvidia GeForce RTX 5080 also features 16GB of GDDR7 VRAM on a 256-bit bus. That should be enough memory for most people, even at 4K, but games are demanding more and more VRAM every year. With the RTX 5090 packing 32GB of VRAM, it would have been nice to see the RTX 5080 adopt the 24GB featured on the RTX 4090, especially since this GPU is being pitched as a 4K graphics card.

Surprisingly, though, Nvidia hasn’t drastically increased the power demand of the RTX 5080 over its predecessor, requiring 360W, compared to just 320W from the RTX 4080. It’s still an increase, but compared to the monumental 575W needed for the RTX 5090, it’s a modest 13% bump.

That modest power increase means that the Founders Edition has to do a bit less work to keep things cool. Despite that, though, the RTX 5080 looks remarkably similar to the RTX 5090, down to the same gunmetal gray colorway and flow-through heatsinks. It’s little wonder, then, that the RTX 5080 peaked at 77°C in my test suite, compared to 87°C from the same cooler on the RTX 5090. Plus, it’s a dual-slot card, compared to the triple-slot design of the RTX 4080 Super, and still cools better, with the last-gen part peaking at 82W in the same test suite, while only consuming 320W of power.

DLSS 4

Rather than monumentally better performance, the big selling point of the RTX 5080 is DLSS 4, or rather the multi-frame generation part of it. In fact, for the most part, many of the improvements Nvidia is making to DLSS are going to be available on older graphics cards. The improvements that carry over consist of a new AI model that’s going to serve as the foundation of DLSS, moving from the Convolutional Neural Network of DLSS 3 and into a Transformer model.

While the change in the foundational model of DLSS probably won’t affect your frame rate in games too much, it should improve image quality, reducing the ghosting and artifacting that could happen, particularly in older versions of DLSS. But that Transformer model will also help reduce the amount of VRAM needed for DLSS in general, which should help people on lower-end graphics cards push to higher resolutions.

Less of a reliance on VRAM is also what allowed Nvidia to implement MFG, or Multi-Frame Generation on the RTX 5080. Just like the version of Frame Generation that launched with the RTX 4080, MFG works by analyzing a rendered frame, along with motion vector data supplied by the game engine, to generate entirely new ‘AI’ frames between each rendered frame. The big difference here is the scale – MFG allows the RTX 5080 to generate up to 3 frames from each rendered frame in supported games.

Just like the previous version of frame generation, this solution can introduce significant latency, but luckily Nvidia Reflex 2 should help combat it. Just like the original version of Reflex, the sequel zeroes out the render queue provided by your CPU, making it provide frame data just as the frame is about to be rendered. But it takes it further, taking your last mouse position, as well as data from previous frames to more accurately guess where you’ll be moving. That’s combined with hardware flip metering, provided by the AMP, which will pace out the frames to provide for maximum smoothness.

However, all of these solutions can only do so much. In order to get the best experience from multi-frame generation, you’re only going to want to enable it when you’re already getting a decent frame rate. I’d advise only enabling it if you’re already getting 60 fps, or 40 fps if you’re feeling extra frisky.

Once all of that is out of the way, DLSS 4 MFG is actually impressive, at least if you have something like a 240Hz 4K display. In Cyberpunk 2077 at 4K, with the Ray Tracing Overdrive preset and DLSS set to performance, the RTX 5080 was able to deliver 59 fps, right where you’d want it to be. Enable 2 x frame gen, like what’s supported on the RTX 4080, and that number jumps up to 104 fps. That’s already impressive, but 4 x frame gen sees that number go all the way up to 183 fps – not enough to fully saturate a 4K 240Hz display, but it’s damn close, and you don’t need to spend $2,000 to get there.

Likewise in Star Wars Outlaws, I saw the frame rate go up from a 61 fps average at max settings with DLSS set to performance, up to 100 fps with 2 x frame gen and 176 fps with 4 x frame gen. That means the frame rate essentially triples.

However, because the underlying frame rate remains around 60 fps, you shouldn’t expect a huge increase in responsiveness. The game will essentially just look smoother. That’s not nothing, but it’s not the same as being able to render the game three times faster.

Performance

There have been graphics cards in the past that have soared above their predecessors, but the RTX 5080 isn’t one of them. But while it doesn’t have the same kind of uplift that the 4080 demonstrated over the RTX 3080, it does still provide the best performance you can get in this price range. In fact, this is the second generation in a row where Nvidia’s xx80 card is a legitimate 4K contender for anyone that doesn’t want to sell a kidney for the 5090.

At 4K, the Nvidia GeForce RTX 5080 is, on average, 11% faster than the RTX 4080, and 8% faster than the RTX 4080 Super, with that performance uplift dropping precipitously at lower resolutions. However, when you add in synthetic benchmarks from 3DMark, that performance differential grows, up to 11% over the RTX 4080 Super and 15% over the RTX 4080. This suggests that this graphics card will get better over time, as new drivers and more demanding games come out.

For this review, I’m using the same comparison data I used for the RTX 5090 last week, along with testing the RTX 4080 and RTX 3080 on Nvidia Game Ready Driver 566.36. The AMD Radeon RX 7900 XTX and 7900 XT were tested on AMD Adrenalin 24.12.1. There was an update to Cyberpunk 2077 that launched halfway through my testing process, but I forced DLSS to run on the older CNN model to ensure the results would be comparable. I did not use DLSS 4 or any type of Frame Generation in any of these tests.

In 3DMark, it’s clear that the Blackwell architecture can do some serious work, especially in games with DirectX 12 and ray tracing. In the Speed Way benchmark, the RTX 5080 gets 8,966 points, compared to 7,519 points for the RTX 4080 Super and 7,258 points for the RTX 4080. That makes for a 23% difference over the RTX 4080, not bad considering the 4080 launched for $1,199 – $200 more than the 5080. In Port Royal, a test that focuses heavily on ray tracing, the RTX 5080 also enjoys a 24% lead over the 4080, but it does lose some ground to the 4080 Super, with that lead dwindling to 19%. A sign of things to come.

Call of Duty: Black Ops 6 is a game that naturally rewards high frame rates, and even at 4K, the RTX 5080 is able to get up to 140 fps at the Extreme preset with DLSS set to ‘Performance’. That’s enough for most high-refresh monitors, but compared to the 129 fps from the RTX 4080 Super and 124 fps from the 4080, it’s not exactly a huge leap. And, compared to the Radeon RX 7900 XTX’s 145 fps with FSR set to ‘Performance’, it’s actually a loss compared to the top-end card from Team Red.

Cyberpunk 2077 is one of the most demanding games on the market, but is getting a lot more accessible as hardware advances. And at 4K with the Ray Tracing Ultra preset and DLSS set to ‘performance’, the RTX 5080 gets 93 fps, compared to 87 from the 4080 Super and 84 from the RTX 4080. That’s just an 11% gen-on-gen improvement from Nvidia in one of its most favorable games.

I test Metro Exodus: Enhanced Edition on the Extreme preset, but without any type of upscaling, because it only supports DLSS, and not FSR. But even without upscaling, the RTX 5080 almost hits the golden standard, netting 58 fps, compared to 53 fps from the RTX 4080. That’s just a 10% generational lead, but still substantially better than the Radeon RX 7900 XTX’s 44 fps in this game.

In Red Dead Redemption 2, the same problem that plagues the RTX 5090 is present here too, as the game is getting a bit old to be testing high-end GPUs on. In the cowboy simulator, the RTX 5080 gets 127 fps at 4K with every setting cranked up to max on Vulkan. Still, the RTX 4080 Super gets 122 fps and the 4080 gets 120 fps. But, hey, at least it’s a 52% lead over the RTX 3080.

It’s always fun testing Total War: Warhammer 3 because it cuts out a lot of the gimmicks used to sell modern graphics cards. No ray tracing, no DLSS, just pure rasterization. And, well, this ends up being one of the 5080’s strongest showings, getting 92 fps to the RTX 4080’s 76. That’s a massive 21% jump in performance, and close to the potential shown by 3DMark.

It’s funny, then, that I have to talk about Assassin’s Creed Mirage, which shows one of the weakest uplifts for the 5080. In this game, the RTX 5080 is just 4% faster than the 4080, but I suspect this is due to a problem with the driver, similar to the issues I saw in this game when testing the RTX 5090.

Black Myth Wukong is another extremely demanding (and gorgeous) game, but it shows similar scaling to Cyberpunk 2077. In this game, the GeForce RTX 5080 gets 74 fps at 4K with the Cinematic preset, compared to 68 from the RTX 4080, making for a 9% difference. Either way, though, this is good enough performance to flip on frame generation and get much higher frame rates out of it – regardless of how ‘real’ they are.

Finally, in Forza Horizon 5, I was surprised to see how much faster the RTX 5080 is than its predecessor. The new card gets 171 fps at 4K with the Extreme preset, compared to 146 fps from the RTX 4080 with the same settings. That’s a 17% performance increase, and makes for the best scaling in a game that has DLSS enabled. This also demonstrates that the RTX 5080 doesn’t quite bottleneck my Ryzen 7 9800X3D quite as much as the 5090 does – at least at 4K.

The Nvidia GeForce RTX 5080 isn’t exactly what I’d call an exciting graphics card, especially compared to previous generations of Nvidia GPUs that just demolished everything that came before. However, while its performance uplift might not be as high as other generations, it’s still an awesome graphics card, especially at 4K. If you’ve been on the fence about upgrading, now might be the time to pull the trigger, especially if you can find the RTX 5080 at its suggested $999 price tag.

Jackie Thomas is the Hardware and Buying Guides Editor at IGN and the PC components queen. You can follow her @Jackiecobra