Nvidia RTX 5080 vs. RTX 4090 — can it dethrone the last-gen king?

When you come at the king, you best not miss. Fortunately, it’s almost a given that Nvidia’s XX80 GPU would beat the XX90 graphics card that preceded it. That’s just how it tends to go. But we’ve been hearing rumblings of the RTX 5080 underperforming expectations since its original unveiling and Nvidia’s praise sweetened with such a heavy dose of AI. Our RTX 5080 review didn’t do much to change that tale.

So, how does the RTX 5080 compare to the RTX 4090? Can it compete? Does it offer a value upgrade for 30-series and 20-series owners looking for the best graphics card they can afford? Let’s find out.

Pricing and availability

The RTX 4090 debuted in October 2022 with an at-the-time record price for a top-consumer graphics card of $1,600. That price bounced around a lot during GPU shortages but ultimately remained relatively constant over the past few years. Stock of the card has all but disappeared at the time of writing, but used models have skyrocketed in price after the mediocre improvements of the RTX 5090. The 4090 is now priced between $1,500 and $2,500 even on second-hand markets.

The RTX 5080 is on sale now, but stock is extremely limited. Its suggested retail price is $1,000, but third-party cards are already up to $1,500, and shortages and scalpers may drive the price up further in the short term.

Specifications

The RTX 5080 is, on paper, a weaker card than the RTX 4090, with around 50% fewer CUDA cores, fewer RT and Tensor cores (albeit from a later generation), 50% less memory, and a narrower memory bus. The RTX 5080 does have a higher boost clock and much faster memory, but we’d be hoping that the new Blackwell architecture does more of the heavy lifting of bridging the gap between these two cards.

The RTX 5080 is more efficient than the last-gen king, too, demanding just 360W, whereas the RTX 4090 could easily pull 450W when pushed hard.

Specs rarely tell the whole story, though, and they certainly don’t cover the new DLSS features that the RTX 5080 supports. Its next-generation Tensor cores let it utilize multi-frame generation, which has the potential to artificially boost frame rates in compatible games by quite some margin.

Performance

Real-world performance testing is always where the GPU rubber meets the gaming road, and unfortunately despite all its potential, the RTX 5080 really doesn’t impress. Its average performance across our suite of games at 4K and 1440p showed it as a very fast card, and 10-15% faster than the RTX 4080 Super that it’s replacing, but falling far short of the RTX 4090 when we don’t make use of the new multi-frame generation feature.

In 1440p the gap between the 5080 and 4090 is lessened, but the 4090 still holds sway, particularly in slightly older games.

Although we didn’t run into any VRAM issues during testing, it’s worth noting that super-demanding games like Alan Wake 2 have been shown to fill as much as 15GB of VRAM when running at full specifications, so we’re already seeing the 5080 potentially miss out on future gaming features because it just doesn’t have the VRAM room as the 4090 does.

Ray tracing performance does see the 5080 almost catch up with the 4090, and pull further ahead of the 4080 Super (as well as obliterating AMD’s alternatives), but the additional RT cores in the 4090 keep it more than relevant as the second-best RT graphics card out there — albeit at distance from the 5090.

The frame generation feature is the only real standout for the RTX 5080, and arguably its only big selling point for any 4080 or 4080 Super owners considering an upgrade. In compatible games, it can double or triple frame rates when enabled, though not all gamers will be willing to make the visual fidelity sacrifices that such a boost to FPS demands. The fewer tensor cores on the RTX 5080 (versus the 5090) means that its performance hit from enabling multi-frame gen is more pronounced, too, so you’re more likely to see the artifacts and other issues frame generation can cause.

Overall, the RTX 5080 is a better card than the 4080 Super, but it can’t quite eclipse the 4090 without its AI-augmented frame generation.

Thermals and sizing

At this level of graphics card with these kinds of power demands, you know they are going to run hard and maybe hot, but with the coolers on these cards, they don’t actually get too toasty. The RTX 4090’s cooler is gigantic and still works really well at chilling down that toasty GPU, but the RTX 5080’s is far slimmer (dual slot!), making it a card that’s easier to fit into smaller form factor builds. That’s not nothing, but it’s a very minor selling point for an otherwise underwhelming GPU.

Begrudgingly, the 5080 is the “better” option

I wish there were enough RTX 4090s out there for everyone to buy at an affordable price because then I’d probably just recommend you buy that instead. It’s still incredibly fast, and though it does demand more power, it has boatloads of VRAM with no signs of saturation on the horizon. But I can’t. Because the RTX 4090 isn’t on sale anymore, and the second-hand options are massively overpriced.

So, where does that leave gamers with around $1,000 burning a hole in their pocket, looking for a GPU upgrade: The RTX 5080 — if you can find it at a good price. If you can’t, one of these might be better. It’s still an improvement over anything besides the 4090 and 5090 (and does get close to the former), and multi-frame generation might not be the game changer Nvidia wants it to be, but in the right setting, it makes a huge difference. There’s a reason we’ve all been using Lossless Scaling so much.

The RTX 5080 isn’t the big, new-gen card we’d hoped it would be, and it is absolutely not worth it if you’re already gaming on a 4080 or 4080 Super. For anyone else though, eh? It’s OK. It might be worth waiting to see whether the RX 9070 XT from AMD gets close at a fairer price, but that’s still a couple of months away. At the very least, don’t pay scalper prices for the 5080. It isn’t worth that.