Nvidia GeForce RTX 5080 review: "the 4K GPU I'd aim for this generation"

I’ve been bonding with the Nvidia GeForce RTX 5080 for about a week now, and I’m convinced it’s more than enough graphics card for most people. Perhaps that conclusion is influenced by the fact paying over a grand for a GPU gives me the ick. It could even be that new Frame Generation tricks are furnishing me with more frames than ever. The one thing I’m certain of is that the 80-class card won’t leave you wanting for 4K performance, even if that does mean embracing its AI abilities.

In an alternate timeline, I’d be poking fun at the Nvidia GeForce RTX 5080 and its $999 / £979 MSRP. There’s still part of me who thinks it’s absurd to pay anywhere near a grand for a GPU, but in a world where the flagship RTX 5090 costs the same as a half-decent used car, it’s now the going rate for the best graphics card options with solid 4K capabilities. Comparatively, the new Blackwell card is also pretty appealing compared to the OG RTX 4080 seeing as that card boldly stepped on stage wearing a bold $1,199 price tag.

The RTX 5080 reassures me that Nvidia has listened to PC gaming heckling over the past few years, and I’m glad we’re not seeing a repeat of the two-tier 4080 fiasco from last time. I firmly believe that 80-class cards are vital, and while I suspect more players will be looking to buy an RTX 5070 for price-related reasons, recommending the RTX 5080 to premium players isn’t going to torment my soul.

Latest Videos From

Specs

Armed with a new Blackwell GB203 GPU, the Nvidia GeForce RTX 5080 benefits from the green team’s next-generation architecture. The chip is packing fifth-gen 336 tensor and 84 fourth-gen ray tracing cores - an uplift compared to the RTX 4080 Super’s third and fourth-gen ensemble. On the CUDA side of things, you’re getting 10,752 cores, which isn’t too far off the 10,240 included with last-gen 80-series Super cards.

As for memory, the RTX 5080 arrives with faster GDDR7 modules while sticking with 16GB VRAM. This leaves a rather large gap between the 32GB included with the flagship RTX 5090 and its 80-series counterpart, and I’m surprised the card didn’t aim to match the Nvidia GeForce RTX 4090 in terms of capacity. Whether you’d actually benefit right now from 24GB instead is another thing, but all it takes is a new PC release to munch more memory than it should to cause system requirement headaches.

The RTX 5080 features a 380W TDP - an increase of around 60W compared to the RTX 4080 Super. That shouldn’t be enough to cause too much power supply upset across the board, but it is worth taking note of if you’re already pushing your PSU to the limit. It’s also not as egregious as the jump from RTX 4090 to RTX 5090, as the new frontrunner guzzles a tremendous 575W.

Design

If you asked me what I thought the RTX 5080 Founders Edition would look like around a year ago, I’d have said it’d be the same size as a full-fat PS2. It’s safe to say I was overjoyed upon carefully lifting the graphics card out of its new eco-friendly prison, as it’s thinner than both the RTX 4080 Super and the OG card.

That’s right, two-slot GPUs are back, baby, and while the RTX 5090 uses the same shroud, it still feels like the right size for an 80-class card. Nvidia has completely won me over with the look of this card, as it looks incredibly slick compared to previous Founders Edition releases and the monstrous-looking custom cards haunting many rigs out there. From its new angled connector that gently steers its 12-pin power port off to the right to its minimized heatsinks, this card is a looker.

Image 1 of 3

Now, I know what you’re thinking, surely the thermals are compromised as a result? Well, Nvidia has actually rejigged its internal design by ditching its previous approach. Traditionally, graphics cards feature a single PCB (printed circuit board), but the green team has separated things into three boards for the 50-series. By isolating the GPU and VRAM modules away from the PCIe connector and video out IO, the company has found a way to better supply the heatsinks with more intentional airflow rather than just having the fans blast air within and hope for the best.

Simply put, Nvidia has modernized its GPU design and conjured up something that’s both smaller and runs cooler. That in itself is praiseworthy, and I’m keen on how it’ll affect the graphics card scene at large. More importantly, it could bury the notion of massive GPUs being mandatory once and for all, and it’ll hopefully lead to future graphics cards getting even smaller.

There is still a 12-pin elephant in the room in regard to Nvidia’s power connector, and while I’ve acclimatized to using a dongle, it’s still a pain. The aforementioned angled connector helps somewhat in terms of case space, but you’ll still need to deal with intrusive wires at the forefront of your rig.

Features

Nvidia and AI are practically synonymous in 2025, so it’s hardly surprising that the tech is an RTX 5080 selling point. You could argue that tools like DLSS and Frame Generation are becoming a mandatory part of PC gaming seeing as the fps boost is extremely hard to ignore, and the green team’s 80-series card has me questioning whether the days of native graphics are numbered.

Not every game supports it, but just like the RTX 40 series, the RTX 5080 is armed with Frame Generation abilities. Only this time, Nvidia has stepped things up a notch with “Multi-Frame Generation,” an RTX 50-series exclusive trick that produces three frames for every single rendered frame. It’s not too dissimilar from the version included with DLSS 3.5, but rather than effectively doubling fps, the tech is boosting fps even further by predicting what’s coming next.

Rather than waiting for Multi-Frame Generation support to come to a bunch of PC games, it can be instead injected into virtual adventures with existing DLSS 3.5 support. That’s pretty neat seeing as there are already 75 romps out there compatible with the RTX feature, and new releases will likely land with support baked in throughout 2025.

YouTube

Watch On

There’s room for debate over whether frames dreamed up by AI are a legitimate performance boost. Frame Generation might feel like magic, but it’s not immune to occasional artifacts and visual weirdness, which act as a Matrix-style black cat moment. I’ll get into my specific experience with the RTX 5080 soon, but the tech in general has come a long way since the early days of DLSS upscaling, and I’m finding less reason to leave it off with every iteration.

Of course, DLSS 4 has more strings to its bow than Multi-Frame Generation, as it boasts all the same abilities as DLSS 3.5. Ray reconstruction, Super Resolution, and Deep Learning Anti Aliasing (DLAA) are all ready and waiting to spruce up your visuals and boost fps, while Nvidia Reflex support is waiting to kick latency issues to the curb.

Perhaps one of the biggest changes to DLSS in its fourth iteration applies to its AI model, as Nvidia is now using Transformer - the same architecture as Google Gemini and ChatGPT. No, that doesn’t mean your graphics card is now a chatbot, nor does it mean you’ll have any additional AI weirdness being injected into your games. Instead, the model improves fidelity through tackling temporal stability, ghosting, and quirks tied to motion. Simply put, you should see less visual shenanigans as a result of generated frames or upscaling.

It's well worth giving a nod to the new Nvidia App too, as the new hub has simplified access to all the green team's GPU settings and features. It pretty much performs the same duties as the old GeForce Experience kit, but provides a more user-friendly way to change visual settings rather than diving into Control Panel. It also serves as centralised place to mess with features like the Nvidia Overlay in addition to the card's AI-powered Broadcast streaming abilities and various other tools.

Performance

Testing the RTX 5080 was a bit of a surreal experience. Don’t get me wrong, I was expecting some modest performance increases compared to the RTX 4080 Super, but Nvidia DLSS 4 effectively poked fun at even my most demanding test scenarios with ridiculous frames during benchmarking. I'll be benchmarking even more games and adding the results to this review soon, but I've already got plenty of frame rates to share with DLSS on and off.

Regardless of your thoughts on AI tricks, it’s hard to argue against using DLSS 4 when it has the ability to pull off such stunts as running Cyberpunk 2077 at 4K with RT Overdrive setting enabled while still hitting 185fps. Effectively, the feature helps the RTX 5080 access path tracing without breaking much of a sweat - something that will even stress out the RTX 4090 armed with vanilla Frame Generation.

I admittedly wouldn’t recommend playing Cyberpunk 2077 on PC without some sort of DLSS assistance. While native performance usually isn’t an issue with ray tracing out of the equation, that all changes with the lighting setting switched on. What I’m trying to say is that while the RTX 5080 hitting just 35fps at 4K with ray tracing switched on looks bad, no one is going to be roaming around Night City without at least Super Sampling switched on. I mean, are you really going to pass up enjoying a 247fps average with Multi-Frame Generation enabled? I think not.

No, that’s not a typo, Multi-Frame Generation really does ramp up fps to triple digits with 4X enabled. You can scale things back to 3X or 2X if you want to stress out your GPU more, but you’re not going to necessarily see a visual difference. I struggled to find anything to grumble about in terms of artifacts while using DLSS 4, and that includes the usual offenders like strange-looking blades of grass and messy mesh fences. The only thing I did pick on was an occasional softness on passing objects when blazing through the streets on a motorbike, but that’s what I get for speeding.

Image 1 of 2

DLSS 4 off (Multi-Frame Generation and Super Sampling)

DLSS 4 on (Multi-Frame Generation and Super Sampling)

Dragon Age The Veilguard serves as another RPG that gets a performance revamp via DLSS 4. Without Multi-Frame Generation, the RTX 5080 pulls off a respectable 66fps with 4K ultra ray tracing settings enabled. Flick the AI switch, and you’re talking 261fps, which is an absurd leap in performance. You’d think that sort of trickery would come with fidelity caveats attached, but after much starring at the finer details of the game’s opening area, I couldn’t find anything out of place. If anything, using frage generation strangely helped combat some distracting glowing gold bowls that should have been elegantly bouncing nearby fire pit glow. Huh.

Okay, let’s step away from DLSS 4 for a bit, as Nvidia’s party trick isn’t going to apply to every scenario. In some of my benchmarking favorites like Hitman 3, I was able to squeeze out higher frame rates using the RTX 5080 compared to the RTX 4080 Super without the use of upscaling. At 4K, we’re taking native results of 164fps compared to 133fps via the Lovelace card. That’s a nice result considering we’ve left DLSS out of the equation, and I’m even pretty impressed with the 10fps increase found with ray tracing switched on (39fps versus 48fps). Naturally, you’ll want to use AI upscaling to hit a much more palatable 133fps instead, especially since the 4080 Super only managed 105fps. But hey, I’m pretty happy with even the native punches being thrown here.

As a control game of sorts, I always put GPUs to work running Total War: Warhammer 3. The RTS romp features no DLSS or ray tracing support, meaning it can paint a more traditional performance picture in terms of graphics card abilities. The RTX 5080 faired pretty well in my usual 4K ultra tests, providing a respectable 87fps average that’s pleasing close to my typical RTX 4090 results (around 95fps).

I’ve been playing a bit of Marvel Rivals lately, so it felt only right to put the RTX 5080 to work running the hero shooter. Normally, I wouldn’t play an FPS game like this at 4K or with any upscaling trickery enabled, but DLSS 4 makes for a ridiculous experience. Without it, you’ll be playing at around 65fps, but Multi-Frame Generation ramps that up to a frankly ridiculous 312fps. That’s more than enough to satiate some of the best gaming monitors out there right now with 4K 240Hz abilities, and if you want to go even faster, you’ll be able to reach 563fps at 1080p.

The takeaway from all those benchmarks is that DLSS 4 and Multi-Frame Generation provide an unprecedented performance uplift. Strip away the AI superpowers, and you’re still looking at a modest native lift that places the RTX 5080 somewhere between the RTX 4080 Super and RTX 4090.

Yes, it would be nice to see the RTX 5080 match or surpass Nvidia’s former flagship, but keep in mind you’re getting a slimmed-down card with a more efficient cooler for the same price as the RTX 4080 Super. It also has comparative efficiency on its side thanks to its 360W TDP, whereas the RTX 4090 guzzles 450W.

As for temperatures, the RTX 5080 keeps things comfortably cool under load, and I rarely witnessed the GPU reach that far past 70°C. More often than not, I’d find the GPU setting at around 65°C, and that includes while handling Cyberpunk 2077 in RT Overdrive mode. Those nice thermals aren’t attached to any noisy fan efforts either, so you won’t have to worry about a chorus of loud hums disrupting your setup zen.

One thing that caught my ear, however, was some light coil whining. You’ll only encounter this audible phenomenon when running games at a higher frame rate than your monitor supports, but it’s something I’ve not heard since using the GTX 970. It’s not quite as aggressive as that old GPU and pretty easy to ignore, but still worth noting if you pick up on even the slightest rig noises.

Should you buy the Nvidia GeForce RTX 5080 Founders Edition?

The Nvidia GeForce RTX 5080 might live in the shadow of its overachieving RTX 5090 sibling, but it’s ultimately the 4K graphics card you should buy this generation. The Founders Edition specifically will furnish your PC with a sleeker GPU with better cooling and a more attractive aesthetic, and while the native performance leap compared to the RTX 4080 Super isn’t revolutionary, DLSS 4 and Multi-Generation could fundamentally change how we view frame rates and performance.

I won’t lie, I’m not entirely comfortable with AI seeping into the PC gaming scene. That’s putting it lightly, and I’d rather avoid a world where I’m washing the dishes while a machine sits and enjoys my Steam library for me (humorous exaggeration). However, as a standalone GPU feature that runs locally rather than relying on a planet-destroying supercomputer in a remote location, DLSS 4 serves to help more players successfully run games with a premium kick, and it helps turn the RTX 5080 into a card that would otherwise be years in the making.

How I tested the Nvidia GeForce RTX 5090 Founders Edition

For a week, I used the Nvidia GeForce RTX 5080 in my main gaming PC, using the graphics card for benchmarking, casual gameplay, and general use. During that time, I took note of aspects like fan noise and temperatures in addition to carrying out separate benchmarks on a variety of games. I specifically tested the next-gen GPU’s 4K Ultra abilities in Cyberpunk 2077, Hitman 3, Dragon Age: The Veilguard, Total War: Warhammer 3, and Marvel Rivals.

For more information on how we test graphics cards and other PC components, check out our full GamesRadar+ hardware policy.

Looking to pick up a next-gen GPU? Check out where to buy the RTX 5090 and where to buy the RTX 5080. Alternatively, swing by the best Alienware gaming PCs if you want an out-of-this-world rig that’s ready to go.