Up until today, the choice of $999 graphics card falls on the shoulders of Nvidia GeForce RTX 4080 Super or AMD Radeon RX 7900 XTX. Both premium offerings provide excellent gaming chops in rasterised titles – 4K60 is easy – though Team Green retains a comfortable lead in ray tracing and arguably has a better upscaling implementation in the form DLSS versus FSR. Now, though, RTX 4080 Super is unceremoniously pushed aside for a new GPU taking the thousand-buck crown. Enter GeForce RTX 5080.
Based on the all-new Blackwell architecture and purposely going big on ray tracing, RTX 5080’s other not-so-secret weapon is multi frame generation. All that said, it still competes with Radeon RX 7900 XTX for mindshare, so let me do the hard work and compare the pair across specs and games.
| RTX 5080 | RX 7900 XTX | Ratio | |
|---|---|---|---|
| Released | Jan 2025 | Nov 2022 | – |
| Codename | Blackwell | Plum Bonito | – |
| GPU | GB203 | Navi 31 XTX | – |
| Process | TSMC 4N | TSMC 5/6 | |
| Transistors bn | 45.6 | 57.7 | 0.79 |
| Die size mm² | 378 | 529 | 0.71 |
| Cores | 10,752 | 6,144 | 1.75 |
| Base clock MHz | 2,295 | 1,929 | 1.19 |
| Boost clock MHz | 2,617 | 2,498 | 1.05 |
| FP32 Boost TFLOPS | 56.3 | 61.4 | 0.92 |
| SM count | 84 | 96 | 0.87 |
| RT cores | 84 (4th Gen) | 96 | 0.87 |
| RT TFLOPS | 170.6 | – | – |
| Tensor cores | 336 (5th Gen) | 192 | 1.75 |
| FP16 Acc TFLOPS | 450.2 | 122.8 | 3.67 |
| ROPS | 112 | 192 | 0.58 |
| Memory GB | 16 | 24 | 0.67 |
| Memory type | GDDR7 | GDDR6X | – |
| Mem. clock Gb/s | 30 | 20 | 1.50 |
| Mem. interface bits | 256 | 384 | 0.67 |
| Mem. bandwidth GB/s | 960 | 960 | 1.00 |
| Board power watts | 360 | 355 | 1.01 |
| Launch MSRP $ | 999 | 999 | 1.00 |
GeForce RTX 5080 is over two years newer than AMD’s head consumer honcho. I know Radeon RX 9070 XT and RX 9070 are slated for a March launch, though all indications point to slower performance than RX 7900 XTX in most games and applications.
AMD’s historical focus on FP32 throughput stands proud against RTX 5080. Navi 31 XTX’s architecture effectively double pumps cores to reach 61.4 TFLOPS, which is around 10% higher than Nvidia.
RTX 5080, however, offers significant ray tracing and AI capability. Going by my calculations, it has over 3x the FP16 Tensor throughput and while ray tracing numbers are not available for AMD, I’m adamant Nvidia is way higher, borne from knowledge of how cards compare in previously tested RT-heavy games.
Nvidia persists with a single monolithic die whereas AMD uses a chiplet architecture for RX 7900 XTX, where main graphics is built on a 5nm process while memory uses a marginally cheaper 6nm node.
AMD also provides a 24GB framebuffer compared to 16GB for Nvidia, though both are serviced with identical memory bandwidth. Nvidia goes narrow and fast – 256 bits and 30Gb/s GDDR7 – while AMD opts for slower and wider. The two GPUs consume almost identical power running full chat, too.
Divergent design philosophies clash in this $1,000 showdown.
Performance
I ran GeForce RTX 5080 and Radeon RX 7900 XTX back-to-back on the 2025 Club386 test system. Keeping all other variables as consistent as possible, the following benchmark numbers can be trusted as a yardstick for evaluating performance across a swathe of applications and games.
Our 7950X3D Test PCs
Club386 carefully chooses each component in a test bench to best suit the review at hand. When you view our benchmarks, you’re not just getting an opinion, but the results of rigorous testing carried out using hardware we trust.
Shop Club386 test platform components:
CPU: AMD Ryzen 9 7950X3D
Motherboard: MSI MEG X670E ACE
Cooler: Arctic Liquid Freezer III 420 A-RGB
Memory: 64GB Kingston Fury Beast DDR5
Storage: 2TB WD_Black SN850X NVMe SSD
PSU: be quiet! Dark Power Pro 13 1,300W
Chassis: Fractal Design Torrent Grey
Application and AI
3DMark Speed Way adds to conventional load by including lots of ray tracing. This is perhaps why RTX 5080 steals a large lead over RX 7900 XTX. 38% is not to be sniffed at.
Removing ray tracing and focussing on pure rasterisation performance has the opposite effect. Benchmark numbers get closer together, suggesting to me that Radeon RX 7900 XTX is still a mighty fine card for older games run at high resolutions with all eye candy turned on. The gap falls from 38% to 23%.
Use a graphics card to render projects in Blender? I send you towards Team Green’s RTX 5080.
Simple AI is a doddle for either card.
This test uses a locally run 13-billion parameter Llama 3.1 model to generate answers to various questions, in the same way programs like ChatGPT work in the cloud. Though a heck of a lot simpler, Llama 3.1 is a great on-device LLM. GeForce RTX 5080 is almost exactly twice as capable.
But what do these numbers really mean? They represent an accumulation of how long it takes for Llama 3.1 to answer your query and how quickly it’s able to build on that with a reasoned answer. RTX 5080, for example, takes 0.28 seconds to start answering and then pushes 150 tokens per second, or how quickly it finishes answering. By comparison, RX 7900 XTX takes 0.82s and returns 108 tokens per second. Looking at them in real time, Radeon doesn’t feel twice as slow.
Gaming
On to the nub of the matter. GeForce RTX 5080 can’t pull away from RX 7900 XTX in Assassin’s Creed Mirage. AMD’s healthy FP32 throughput and generous memory bandwidth and Infinity Cache hold it in great stead. Though it loses, I mark this as an indirect win for an older design.
Another rasterisation-heavy title, another closely fought contest. You can look at this in two ways. The first is to think Nvidia deliberately leaves raster performance on the table – RTX 5080 ROPS count is much lower than RX 7900 XTX – and goes after RT and Tensor/DLSS; the other is to applaud AMD for making a card this competitive in 2022.
Throwing in a generous dollop of ray tracing separates the pair at each resolution. Nvidia usually argues its GPUs are more forward looking in nature. I tend to agree.
It takes up until UHD for Nvidia to squeak past AMD.
This matchup is closer than I originally suspected. No one’s going to complain at 4K150+ from either card, though.
Here’s where it gets interesting. I run Cyberpunk 2077 with Ray Tracing Overdrive, which is a setting that includes GPU-bashing path tracing.
In many ways, this is an ideal scenario for GeForce RTX 50 Series cards because it very much plays to their architectural strengths. I know this because RTX 5080 is twice as fast as RX 7900 XTX at each resolution, but I question the usefulness of either at UHD.
DLSS
Adding DLSS and FSR3 Quality upscaling in the mix elevates frame rates markedly. But here’s the thing; Nvidia remains twice as fast.
Sticking on each card’s best-performing frame generation reinforces Nvidia’s dominance with this type of technology. All that said, it’s worth pointing out higher framerates at UHD aren’t that useful a metric. The reason for this is DLSS/FSR work best when the underlying frame rate is >30fps, as one cannot magic fluidity out of a slideshow, no matter what the on-screen frame rate says.
| Game | RX 7900 XTX to RTX 5080 % uplift at 4K |
|---|---|
| Assassin’s Creed Mirage | 5.8 |
| Final Fantasy XIV Dawntrail | 7.7 |
| Forza Motorsport | 47.5 |
| Mount and Blade | 6.1 |
| Rainbow Six Extraction | 9.0 |
| Cyberpunk 2077 DLSS + FG | 270 |
I’ll preface 4K results by saying they’re absolutely dependent upon game. Those which mainly use rasterisation – four of six here do – see single-digit performance increases from RX 7900 XTX to RTX 5080. In that context, I’m more impressed with AMD than with Nvidia, particularly as the former provides a 50% larger frame buffer into the bargain.
Never has GPU evaluation been so fraught with methodology. Cranking up ray tracing gives RTX 5080 a 47% lead, while the magic button that is multi frame generation takes performance to a completely different level, though it may not experientially feel that much faster or better.
If you’ve found this comparison useful, please do read the following head-to-heads.
GeForce RTX 5080 vs. RTX 5090
GeForce RTX 5080 vs. RTX 4080 Super
GeForce RTX 5080 vs. RTX 3080
Vitals
What do you know, near-identical system wide power consumption.
And both coolers perform well in keeping them cool.
Nvidia’s Founders Edition is plain better at matching up low temperatures with low noise.
Conclusion
It’s well known in gaming circles that AMD is leaving the $999+ graphics card segment to Nvidia for the time being. A wasted opportunity in my book as existing Radeon RX 7900 XTX proves to be a fine GPU even in 2025. My numbers suggest it performs closely to GeForce RTX 5080 in rasterisation and light ray tracing scenarios, which is some feat for an architecture over two years old and one carrying far greater card memory.
Nevertheless, RTX 5080 cracks RX 7900 XTX when oodles of ray tracing are added to the mix – see Forza Motorsport and Cyberpunk 2077 as proof points – and I contend Nvidia has comfortably superior upscaling technology in the form of DLSS run with Transformer models. Then there’s multi frame generation to consider, which is a technology that AMD has no current answer to. It’s available on 75 games and apps on launch, implemented by either native support or override in the control panel. It’s what I call a Big Deal™.
Putting these parts together leads to me to the inevitable conclusion that RTX 5080 is a better graphics card than RX 7900 XTX if both cost the same money. Muddying waters, AMD’s finest is currently available for $870 or so, and I expect partner RTX 5080s to hit shelves north of $1,100. If pricing holds true, the decision becomes a lot tougher.
My key takeaway is there’s no bad card at the premium end of the market. GeForce RTX 5080 carries on the good work established by numerous previous generations, though I’d really appreciate a future Super variant with 24GB of onboard memory. Come on, Nvidia, you can do it.
GeForce RTX 5080
“Gear up for game-changing experiences with the Nvidia GeForce RTX 5080 and AI-powered DLSS 4.” Read our review.
Radeon RX 7900 XTX
“AMD has a credible premium product with Radeon RX 7900 XTX.” Read our review.
