Nvidia RTX 5080 review: it’s slower than the 4090

Last week, I was optimistic. This week, I’m worried. Nvidia followed up the most powerful GPU it’s ever made, the RTX 5090, with this — the RTX 5080. It’s half the price, and it offers around 70% of the performance. That should be a good thing, but the RTX 5080 falls short when compared to Nvidia’s last-gen offerings that still rank among the best graphics cards.

Only one GPU into Nvidia’s new generation and last-gen cards are offering better performance. That’s not a great start to the generation, especially considering the $1,000 price tag that Nvidia is asking for with the RTX 5080. If you’re coming from something several generations old, the RTX 5080 is your best bet at this price. But for enthusiasts with recent hardware looking for a true upgrade, you won’t find much here.

Nvidia RTX 5080 specs

The RTX 5080 is the second GPU from Nvidia’s Blackwell generation that we’re seeing, and it starts to paint a full picture of the company’s latest architecture. Unlike previous generations, Nvidia’s RTX 50-series GPUs don’t have a significant node advantage. It’s still using the TSMC N4 process and instead focusing on specs like memory, core count, and power draw to find big performance increases.

That means the specs tell a decent story about what to expect. On the core front, you’re getting a mere 5% boost in CUDA, ray tracing, and Tensor cores. However, Nvidia is now using an AI Management Processor (AMP) to schedule tasks across these different cores, which likely contributes a bit to the generational performance increase.

Don’t worry; Nvidia is offering more than a 5% boost in performance. That largely comes on the back of the memory interface and the increased power draw. You get 16GB of memory on the RTX 5080, just like the RTX 4080 and RTX 4080 Super, but Nvidia is using new GDDR7 memory. Even on the same 256-bit bus, Nvidia is delivering around a 25% increase in memory bandwidth.

On the power front, Nvidia increased the power draw by 12%, going from 320 watts on the RTX 4080 Super to 360W. You’ll still be fine with a 750W power supply, just like the previous generation, but at least some of the performance improvement here comes from feeding the cores with more power.

4K gaming performance

The key takeaway here is that the RTX 5080 is slower than the RTX 4090 at 4K, and not by an insignificant margin. Across my test suite, it was nearly 9% behind the RTX 4090. The expectation is that Nvidia should be able to match the performance of its last-gen flagship, and hopefully exceed it. For context, the RTX 4080 delivered 36% better performance than the RTX 3090 on average, and 22% better than the RTX 3090 Ti.

Looking at the gen-on-gen perspective, Nvidia is delivering exactly what it promised. On average, the RTX 5080 is 18% faster than the RTX 4080 Super, and by extension, the base RTX 4080, as well. Despite more disappointing performance than the RTX 5090, the RTX 5080 is delivering a significantly better value — it’s half the price but delivering about 70% of the performance on average.

Digging into individual games, you can see why the RTX 5080 falls short of Nvidia’s last-gen flagship. The margins are thin between the RTX 4090 and RTX 5080 in Assassin’s Creed Mirage and Red Dead Redemption 2, with the last-gen card coming out on top. And it takes an absolute beating in Call of Duty: Modern Warfare 2. 

In a good chunk of games I tested, however, the two GPUs were in lockstep. That was the case in Dying Light 2, Horizon Zero Dawn Remastered, and Forza Motorsport. This is the experience Nvidia should be delivering, at minimum. I expected to see some wins for the RTX 5080, even on thin margins. But test after test, they just never showed up — minus one exception; more on that soon.

It’s not all bad news, though. Although the comparison to the RTX 4090 is disappointing, the generational improvements still matter. Cyberpunk 2077 is a good example of where those generational improvements matter. The RTX 4080 Super and RX 7900 XTX can’t manage a stable 60 frames per second (fps) at 4K without upscaling help, but the RTX 5080 can — and it can do so comfortably.

That’s not the only example, either. Black Myth: Wukong is another game where the RTX 4080 Super can’t reach that 60 fps mark but the RTX 5080 can. This is the only game I tested where the RTX 5080 provided an improvement over the RTX 4090. I suspect that has something to do with this game’s use of Unreal Engine 5. It’s the only UE5 game in my test suite, and the RTX 5090 also saw an outsized improvement.

You have to take the good with the bad, though, and there are definitely some bad points here. Returnal is probably the worst example. Not only is the RTX 5080 15% slower than the RTX 4090, it’s a mere 10% faster than the RTX 4080 Super. Regardless of the perspective you look at, the RTX 5080’s performance in this game is not good.

I’d tell anyone to buy an RTX 5080 over an RTX 4080 or RX 7900 XTX given that they all have the same list price. It’s much more difficult to justify if you can get one of those GPUs for less. At $800 even, the RTX 4080 or RTX 4080 Super gets you a much better bang for your buck. Unfortunately for buyers, the RTX 4080 Super is hard to come by at $1,000, much less below it. I suspect the RTX 5080 will be caught in a similar situation, so pay close attention to the price shifts as the RTX 5080 rolls out.

1440p gaming performance

Down at 1440p, the situation is largely the same, though with a bit tighter margins. The RTX 5080 is 3% behind the RTX 4090 and 17% ahead of the RTX 4080 Super. You’re pushing toward a CPU bottleneck at 1440p, and you might run into one if you push upscaling hard. But you can still get away with using the RTX 5080 at 1440p assuming you have an upgrade to a 4K monitor in your near future.

In specific games, the performance picture changes just a bit. Alongside Black Myth: Wukong, where the RTX 5080 leads the RTX 4090 by 12%, Nvidia’s latest also posted a higher result in Dying Light 2 and Horizon Zero Dawn Remastered. In the latter, you can see this wall that all of the last-gen GPUs run into around 125 fps which both the RTX 5080 and RTX 5090 overcome without issue.

If the RTX 5080 offered this level of performance across games, especially at 4K, it’d be an easy recommendation. That’s unfortunately not the case. Too many titles look like Red Dead Redemption 2, Returnal, and Call of Duty: Modern Warfare 2 where the RTX 5080 is clearly behind the RTX 4090 and only provides a slim uplift over the RTX 4080 Super.

There are some games where the RTX 5080 squeezes out a win, such as Cyberpunk 2077. It’s only a win by technicality, but a win nonetheless. More importantly, I saw a 25% improvement with the RTX 5080 over the RTX 4080 Super in this game, which is quite a generational improvement. It’s just a shame that this level of performance isn’t available across the board.

Flipping over to games like Forza Motorsport and Assassin’s Creed Mirage, the dynamic with the RTX 5080 and RTX 4090 remains the same, but the generational uplift over the RTX 4080 Super is smaller.

Nvidia is clearly prompting a readjustment of expectations here. In previous generations, I’d expect the RTX 5080 to beat the RTX 4090, but Nvidia’s flagships continue to become larger, more expensive, and more power-hungry. The RTX 5080 is half the price of the RTX 5090, after all, placing a ton of distance between Nvidia’s flagship and the next step down.

The RTX 5080 certainly looks better at 1440p than it does at 4K, but when you’re spending $1,000 — and likely much more than that — on a graphics card, you want 4K gaming performance. The margins are better at 1440p, but they’re still not great, making the the RTX 5080 a tough sell regardless of what resolution you look at.

1080p gaming performance

At 1080p, the RTX 5080 finally manages to outclass the RTX 4090, but it’s a bit of a hollow victory. You shouldn’t spend this much on a graphics card to play at 1080p, and the thin win for the RTX 5080 largely comes on the back of games that become bottlenecked by the CPU down at 1080p.

I won’t retread the specific games here again; there really isn’t much of note. You’ll see frame rates well into the triple digits across most games, and with relatively thin margins between the RTX 5080, RTX 4090, and RTX 4080 Super. The results are different between the three GPUs, though the practical experience of using each of them isn’t.

Ray tracing and DLSS 4

Ray tracing isn’t a strong point for the RTX 5080. Nvidia still dominates when it comes to ray tracing performance, especially at this price, where it’s the only option. But when it comes to the gen-on-gen improvements, the RTX 5080 lags.

Dying Light 2 is a good example of that. Without ray tracing, the RTX 5080 managed to match the RTX 4090 at 4K and even exceed it at 1440p, but it’s clearly behind when you toggle ray tracing on. The RTX 5080 is still 17% ahead of the RTX 4080 Super at 4K, but there’s something about the RTX 4090 — be it higher power draw, more CUDA cores, or a larger memory interface — that allows it to scale higher in a way that the RTX 5080 can’t.

Returnal is caught in a similar situation. Both of these titles are light on ray tracing by today’s standards, and Nvidia isn’t providing the gen-on-gen improvement in them as it once was.

Unfortunately, that carries over even into Cyberpunk 2077, which was once a strong point for any Nvidia GPU review. I’m using the Ultra RT preset, but you could utilize the RT Overdrive preset (path tracing) with a little help from DLSS 4, as well. Without going to full path tracing, the RTX 5080 looks as disappointing as it does in Returnal and Dying Light 2. 

With full path tracing in Black Myth: Wukong, the RTX 5080 is certainly better, offering identical performance to the RTX 4090. Still, this isn’t a great performance. As mentioned above, this is the only game where the RTX 5080 cleanly beats the RTX 4090 at 1440p and 4K, and when you toggle on ray tracing, that performance lead slips. Nvidia isn’t providing a huge uplift in ray tracing performance, and that factors into weighing its overall average down.

How do you combat the performance dip? DLSS 4. Like the RTX 5090, you get DLSS Multi-Frame Generation on the RTX 5080. That allows you to generate up to three extra frames for each frame the GPU renders, and at launch, it’s supported in 75 games. DLSS 4 is very good — you can see an example of how good in Marvel Rivals above — but in the case of the RTX 5080, it mainly serves to make up for performance the card should already have.

Although DLSS 4 is great, it doesn’t do enough to justify spending $1,000 on a new GPU, especially if you already have an RTX 40-series GPU with DLSS Frame Generation or third-party tools like Lossless Scaling. DLSS 4 is certainly the best version of multi-frame generation I’ve seen, but it’s not worth spending $1,000 or more on alone.

A sign of the times

The RTX 5080 left me worried. The stakes aren’t that high for this graphics card; Nvidia has carefully controlled last-gen stock, so when it comes to the price you’ll pay and the performance you’ll get, the RTX 5080 is the right choice in this price bracket. But the RTX 5080 sets a negative tone for the rest of the generation.

Nvidia is using the same node with largely similar specs as the previous generation, and that shows in the performance. If this is what Nvidia is delivering at the same price as its last-gen offerings, I’m worried about what the RTX 5070 Ti and RTX 5070 will look like considering they’re $50 less than their last-gen counterparts.

That’s a conversation for a different day, though. For the RTX 5080, it’s the best GPU you can buy in this price bracket — but you’ll likely be gritting your teeth and wishing you could get more for your money.