There are a few different headlines you should expect to see in discussion of Nvidia’s latest graphics card, the GeForce RTX 5080. So let me cut to the chase and summarize.
First up: the 5080 is not a card that has evangelism-worthy performance gains over its predecessor. In terms of the raw rendering power improvement it sits firmly in the good-not-great category. But that’s okay because of point two: the 5080 offers the best ‘raw value’ in performance terms not only of this generation of cards so far – but one of the best we’ve seen in several years.
These two concepts might seem vaguely at odds, but they really aren’t. The performance gains compared to last generation’s RTX 4080 Super, this card’s direct predecessor, are on average somewhere between 10% and 15% – while retaining the same price point. This is a modest gain compared to what we’ve seen in the past, but it’s also a monetary bang-for-buck value demonstrated. But furthermore, the new generation 50 series GPUs have access to the full suite of RTX features that make a major difference to games – more on which later.
One could dismiss this as marketing fluff, and I’m sure some will. As embargoes lift I expect some reviews to lead with things like ‘weaker than the 4090’, which to be fair the 5080 is. But I think this is an upgrade to consider with some nuance. If you already have a 40-series GPU, you probably aren’t intending to drop near enough four figures on an incremental upgrade. This card probably isn’t for you. But if you’re the sort of person shopping in that price range with a card that’s a couple of generations old, the 5080 becomes a strong proposition – and quite possibly the best card on the market in its price bracket, at least without delving into the quagmire that is the used market.
Crucially, pretty much all of the upgrades as detailed in my RTX 5090 review make their way over to the 5080 as well. There’s the new form factor, better heat dissipation with snazzy liquid metal and a custom cooler design, and just generally a slick-looking design on the Founders Edition cards that I really rather love. Like the 5090, the 5080 shrinks compared to the previous generation – dropping from a hefty double-thick three-slot card to a more traditional two-slot design. This is a small change – but I actually think it’s rather important for this card.
In real terms, if you’re a normal person, I think the 5080 is the highest-end graphics card you should consider. It’s true that the 90s are the ‘top of the line’ – but the power level in those cards are for sickos, freaks, and people who need to do super challenging work like 3D rendering, complicated video editing or streaming setups, and so on. The same delineation was true for the 4080 and 4090 – but both were heaving three-slot cards. On the 90, the smaller form factor is a nice-to-have. But on the 80 cards, I actually think it’s important.
To my mind, the ‘average gamer’ doesn’t tally with people who have cases and setups built to hold such a beast. By bringing the physical size of the GPU down and by keeping the price under that $1000 mark, the 5080 presents once more as something more suitable for all gamers who have a grand to burn and want the best performance for that money they can possibly get.
People in that consumer bracket can as previously mentioned expect a gain of around 10-15% in raw rasterization performance – that is to say, how well the GPU renders your games without any additional features helping out – when compared to last year’s 4080 Super, which shipped at the same price. To back that up, let’s talk broadly about a few benchmark games.
If we hop into Cyberpunk 2077 with its settings maxed out but with no ray tracing or frame generation technology, I experienced more or less an exact 15% increase on the frames generated by the 4080 Super. Hop over to Alan Wake 2 and it’s a similar story, though with slightly less potency, offering a boost of 10%. Star Wars Outlaws is the same. The performance change is quite linear and predictable across all games be that at 4K or 1440p – almost always broadly the same percentage lift, always matching that initially set expectation.
Each of these benchmarks underlines one other thing: despite that power boost over the 4080 Super, the 5080 remains around 10% behind the 4090. This may disappoint some. In previous generations, we have frequently seen the new generation of one Nvidia card match the performance of the last generation’s next card up. For instance, I loved that the 4070 had a power level similar to or in advance of the 3080 – that helped to make it feel like a full ‘generational leap’. It’s undeniable that the 5080 doesn’t accomplish that, with the 4090 leading it to the tune of up to 10% on average – but does it matter? Second-hand 4090s are still costing almost double what this card does on the second-hand market – so is the comparison even fair?
This power boost might seem modest, but because the price has remained ‘frozen’ at $999, in terms of value for money it still works out pretty well. Our buds at DigitalFoundry have put together a pretty fascinating chart, for instance, that shows that the 5080 is the single best GPU for the number of frames it generates on average per each dollar spent – the second place being the 3080, a card that at this stage is pushing five years old. Mercifully, power consumption remains quite close to the 4080 Super also – meaning it’ll cost about the same to run and avoids the eye-watering leap in running costs we saw in the 5090.
In comparisons like this we have to step back from raw rendering and frame rate percentage increases from one generation to the next and take in the whole picture. That whole picture is more flattering than those first numbers would first present.
There’s more than just the modest increase in raw power that this generation’s Blackwell architecture unlocks, however. Each 5080 of course packs in the fifth generation of the Tensor cores that unlock the ‘RTX’ functionality, which in turn opens the door to a variety of software solutions that can transform games and productivity.
As with previous generations, the charge on this front is led by Nvidia DLSS. The latest version is DLSS4, and is a key feature of the 50 series – as is the new Multi-Frame Generation (MFG) technology. This is the comeback to any perceived shortcoming in raw rendering upgrades: Nvidia’s technology beyond that feels quite unassailable, and the newest cards get the best and most powerful iterations of this tech.
If you’re unfamiliar, DLSS stands for Deep Learning Super Sampling. It’s essentially a system that uses AI learning in order to help the graphics card to deliver a higher image quality without sacrificing frame rate. It’s all a bit of smoke and mirrors – but it’s a rock solid bit of fakery that works brilliantly.
Essentially, DLSS allows the GPU to render the game at a lower resolution than whatever your display throws before your gawping eyeballs. A lower resolution makes it easier to render more complicated things, like real-time ray tracing or just very taxing games, at a high frame rate. But DLSS then steps in and upscales the generated images – so you might render the game natively at something like 2K, but then DLSS will step in to give you a crisp 4K image. As the technology has improved, signs the image has been upscaled have diminished to be near imperceptible, while other potential pitfalls like lag have been offset by other innovations like Nvidia’s Reflex software.
Multi-Frame Generation takes this to the next level. As well as upscaling an image, it allows the GPU to use AI to essentially create whole new frames out of nowhere, extrapolating from the frames it’s rendering in real-time. These ‘fake’ frames can make a game appear smoother and easier on the eye, though they won’t affect a game’s responsiveness. MFG is an all-new technology for this generation, and I’d say its primary utility is in helping to get games up to 120 frames per second for 120hz displays.
Both technologies work well, and serve a vital purpose. Even at 4K this card crushes Cyberpunk without ray tracing, offering up something silky and acceptable to even the more picky player. But flick on ultra level ray tracing in Cyberpunk and you’ll see that at 4K this is a card that still struggles to hit a solid 60fps with that technology enabled. Enable DLSS, however, and hitting 60fps with top-line ray tracing is a breeze.
MFG can even put 120fps in sight on such settings – which is just proof that this technology is the future. At this stage, however, it’s hard to truly judge MFG – it’s impressive, but the limited suite of games available to test (which will expand to around 75 at launch) leaves me wanting to see it used more. Then again, I had the same skepticism about DLSS years ago – and now I use it without question.
A lot of hardcore graphics nerds are completely reasonably going to tie themselves in knots over this. Do fake frames count? A lot of these people are trapped in the tradition of primarily being only interested in raw processing power – which I totally understand. But at this point this technology is an indelible part of the make-up of graphics card buying decisions. More and more games are supporting this sort of technology; it’s likely to be a vital component of the console market going forward, too. It’s here to stay.
When you consider that, one key element is clear: anyone buying a GPU like the 5080 is not only buying the raw compute power that Nvidia’s new Blackwell architecture unlocks – they are buying into the software assistance the card unlocks for high-end games. Like it or not, this is now a core part of Nvidia’s proposition. Sometimes those raw compute leaps are going to still be jaw-droppingly large – but not always. The 5080 is an example of that – where a modest but reasonable GPU upgrade in raw terms is heavily augmented by the other features the card brings to the table.
I think, in honesty, by this point in the review you know where you stand. You know if the concept of a 15-20% boost in raw power is acceptable or a huge turn-off to you, you know how you feel about your current setup, and you know how likely you are to engage with DLSS4 and MFG. What I can say from my testing is that all these things work astonishingly well. The new ‘Transformer’ model for DLSS is a huge upgrade and a particular favorite of mine – though that is also available on 40-series cards.
That perhaps sums it up, though. If you’re cruising on a 40-series GPU, many of the best features of the 50 series are available to you already via software updates. To those, this generation might prove a bit disappointing, a bit of a middle step. To anyone upgrading from something like a 30 series or older, the 5080 is a solid proposition at $1000 – a tantalizing upgrade that’ll offer far more of a noticeable performance boost and new software tech you otherwise wouldn’t have access to.
In short, the 5080 is a strange graphics card. If you want the best GPU available up to $1000… this is the one to buy. But in a sense, in the current market, it is also the only one to buy in its class. That’s one bottom line. The other is that this doesn’t hold the usual power gain you’d expect from a generational shift – but software goes a long way to soften that fact. I expect it’ll be divisive – and I expect you already know if you want one or not.