Verdict
There’s no denying Nvidia’s GeForce RTX 5080 delivers some seriously impressive frame rates at 4K. All you need to do is grab a solid PSU and one of the best monitors you can find, and you’ll have a seriously good time with any title you choose to throw at it.
-
Impressive 4K performance -
Twin-slot shroud on FE is a dream -
Multi-frame gen is solid
-
Availability is non-existent -
Not a huge leap compared to the 4080 Super -
MFG can be controversial
Key Features
-
Review Price: £1000 -
Blackwell Architecture
Nvidia’s newest GPU architecture, Blackwell, delivers some impressive AI clout while leaning heavily on TSMC’s latest 4NP manufacturing process. With 10,752 CUDA cores, in this config, the RTX 5080 delivers some serious performance in game -
16GB of GDDR7
Nvidia’s utilizing Samsung’s latest GDDR7 memory modules, and these things absolutely rip, delivering some of the highest speeds we’ve ever seen for VRAM. The 5080 delivers 960GB/s of total memory bandwidth -
DLSS 4
A new generation of cards means new DLSS. Nvidia’s moved to a transformer-based model to improve overall image quality under upscaling and introduced Multi-Frame Generation too, allowing you to add 2, 3 or even 4 additional frames in-between, drastically boosting framerate
Introduction
This may well be Nvidia’s most controversial graphics card launch to date, as much as it pains me to say it.
The Nvidia GeForce RTX 50 series has had one heck of a bumpy debut, with stock being conspicuous in its absence and certain cards from particular manufacturers lacking the full complement of componentry they’re meant to have, yet still going on sale.
That said, the RTX 5080, in isolation, remains an impressive card. At least on the surface. At its core, Nvidia’s 50 series features the all-new Blackwell architecture. Built off the back of TSMC’s 4NP manufacturing process (effectively a refined 5nm node), the RTX 5080, on the surface at least, isn’t that different from its predecessor, the 4080 Super.
Certainly outside of raw specs anyway. There’s only been a modest bump in the total number of components on the thing, with the major performance increase actually coming from clock speed and supercharged memory, more than anything else.
Yeah, it might not be topping the best graphics card of 2025 list, at least not yet, but there’s more to this GPU than meets the eye.
Specs
Core specs, then. As I already mentioned, the jump up from the RTX 4080 Super to the RTX 5080 has been rather modest. Despite utilising TSMC’s tweaked 4NP manufacturing process, the overall footprint of the GPU has shrunk slightly. Not by much, admittedly, 0.6 mm² to be exact, but enough to rein in any major hardware density increases.
Total CUDA cores, TMUs, ray tracing cores, and Tensor cores have all increased in number by exactly 5% total, and that’s it. ROPs remain identical. Not a huge generational leap by any means. Additionally, the advertised boost clock has also increased from 2550 MHz to 2620 MHz (in testing, it peaks at around 2835 MHz in my setup).
| RTX 5090 – 32GB | RTX 5080 – 16GB | |||||
| GPU | GB202-300 | GB203-400 | ||||
| INTERFACE | PCIe 5.0 x16 | PCIe 5.0 x16 | ||||
| DIE SIZE | 750 mm2 | 378 mm2 | ||||
| LITHOGRAPHY | TSMC 4NP | TSMC 4NP | ||||
| TRANSISTORS | 92.2 Billion | 45.6 Billion | ||||
| CORES | 21,760 | 10,752 | ||||
| BOOST CLOCK ADVERTISED | 2,407 MHz | 2,620 MHz | ||||
| BOOST CLOCK RECORDED | 2,902 MHz | 2,835 MHz | ||||
| MEMORY | 32GB GDDR7 | 16GB GDDR7 | ||||
| MEMORY BUS | 512-bit | 256-bit | ||||
| MEMORY BANDWIDTH | 1,792 GB/s | 960 GB/s | ||||
| TGP | 575W | 360 W | ||||
The big winner here for Blackwell is the inclusion of Samsung’s latest GDDR7 VRAM. The speeds capable on this are just unparalleled right now, with the 256-bit bus on the RTX 5080 delivering a whopping 960 GB/s of total bandwidth, giving it some serious clout at 4K.
Although again, there aren’t any total capacity increases here either, so you’re still stuck with that 16GB. TDP has gone up too, increasing from 320W to 360W, and the MSRP is set at the same $1000/£1000.
Buy GeForce RTX 5080 Graphics Cards: Amazon / Currys / Laptops Direct / Overclockers / Scan
Nvidia’s really banking on its progress with DLSS 4 for these cards. That’s the big seller here. Particularly as it’s transitioning away from that convolutional neural network to a transformer-based model instead.
In part, this is to help shift from those often blurry edges and motion sickness that some are susceptible to in the earlier model, particularly with upscaling and frame generation. It’s the latter that’s seen the biggest improvement, with options now available to double, triple, or even quadruple the number of frames generated, significantly boosting framerates in the process.
Test Setup
I’ve spent the last two to three months testing Nvidia’s RTX 50 series cards in all manner of builds and setups. Here, I wanted to do something a little different by entirely revamping our testing suite to really push these graphics cards to the limit and better understand how they operate under load.
Throughout the testing process, I’ve monitored power draw and system temperatures utilising a mixture of plug-in power meters and HWMonitor.
For the tests themselves, I’m using a mixture of synthetic and “real-world” benchmarks from game engines. I’m also using a number of AI tests (including inference testing and image generation), plus some custom data points based on price-to-performance ratios and average frames per second across resolutions.
For our gaming tests, each primary resolution is benchmarked, with a total of five games, each run three times, to net a median result for both minimum and average frame rates.
The games in question are Cyberpunk 2077, Black Myth Wukong, Metro Exodus Enhanced Edition, Final Fantasy XIV, and Total War: Warhammer 3 (utilising the mirrors of madness benchmark). All games are set to the maximum preset available, with no changes; the only exceptions are Black Myth Wukong, which has its ray tracing features disabled and the graphical profile set at Ultra, alongside Cyberpunk, which has DLSS disabled.
Additionally, to get a better understanding of how these cards perform with both DLSS and Frame Generation active, I’ve also done an additional four sets of tests with Cyberpunk and Black Myth Wukong (this time with ray tracing enabled), with AI upscaling only runs, plus AI upscaling and frame generation too.
For Nvidia cards, that of course means we’re utilising DLSS; for AMD and Intel, that’s purely FSR, not Intel’s XeSS.
- CPU: AMD Ryzen 9 9900X
- RAM: 32GB (2x16GB) Team Group T-Create Expert DDR5 @ 6000 C34
- Motherboard: ASUS ROG Strix X870E-E Gaming WiFi
- CPU Cooler: Tryx Panorama 360mm AIO
- Cooling: 10x NZXT RGB Duo 120mm fans
- PSU: 1500W NZXT C1500 2024 80+ Platinum PSU
- SSD 1: 2TB Crucial T705 PCIe 5.0 M.2 SSD
- SSD 2: 4TB Crucial T500 PCIe 4.0 M.2 SSD
- Case: NZXT H9 Elite
As for my hardware picks, you can see those above. Right now, the battle between Intel and AMD for CPU dominance and the hunt for the best gaming CPU is incredibly tight.
Although an X3D chip might have been a better pick, given its 3D cache solution has fairly mixed results depending on the title, I opted to instead stick with the stock AMD Ryzen 9 9900X, as it represented a good balance between single and multi-core performance without sacrificing too heavily on clockspeed.
Gaming Performance
- 60+ fps at 4K
- Not a huge generational upgrade
- Bottlenecks CPUs at lower resolutions
Look, let’s be real here; we all knew this card was going to seriously deliver at 4K, and, well, it really does.
Average frame rates across all of our testing titles landed at 61.56 fps, and that’s a figure dragged down by some pretty aggressive testing on both Cyberpunk and Black Myth. Total War’s Mirrors of Madness benchmark landed a solid 64.3 fps on average, and Final Fantasy XIV landed 111 again at 4K, with not a jot of ray tracing or DLSS in sight.
Metro Exodus, with the profile set to extreme, glided in at 58 fps, Black Myth at 40 fps (without ray tracing, it’s just an aggressive game), and Cyberpunk managed 34.7 fps as well.
| AVERAGE ACROSS ALL 5 TITLES | – FPS – | ||
| 1080p – Average FPS | 120.93 | ||
| 1080p – Minimum FPS | 69.52 | ||
| 1440p – Average FPS | 98.63 | ||
| 1440p – Minimum FPS | 60.01 | ||
| 4K – Average FPS | 61.56 | ||
| 4K – Minimum FPS | 43.49 | ||
All good, impressive numbers. Comparing it to last gen, though, is where the problems begin to arise. The performance difference amounts to around 12.9% according to my figures. That’s an improvement, sure, but in reality, we’re talking about a difference in average framerate of about 6.5 fps.
Now, yes, it has only been one year since the Super’s launch, but it feels massively lacklustre for a generational jump. That only gets progressively worse as we go down the resolution stack as well. At 1440p, the difference is only 4.4% in average frame rates; at 1080p, it’s not even worth reporting—it’s that low.
In the synthetic testing, things were (perhaps unsurprisingly) a lot more positive for the latest generation of card, with the RTX 5080 seeing significant performance bumps compared to last gen as well, but given these aren’t exactly based on real-world gaming, they don’t really serve as much of a comfort. You can see all the benchmarking data below.
Gaming Benchmark Data
| 1920 x 1080 – GAME TESTING | ||||
| Cyberpunk 2077 | Average Framerate – Runs | 93.0 | 92.8 | 92.9 |
| Cyberpunk 2077 | Minimum Framerate – Runs | 79.5 | 78.8 | 78.8 |
| Cyberpunk 2077 | Average Framerate – Overall Average | 92.9 | ||
| Cyberpunk 2077 | Minimum Framerate – Overall Average | 79.0 | ||
| Black Myth Wukong | Average Framerate – Runs | 82.0 | 84.0 | 83.0 |
| Black Myth Wukong | Minimum Framerate – Runs | 66.0 | 64.0 | 65.0 |
| Black Myth Wukong | Average Framerate – Overall Average | 83.0 | ||
| Black Myth Wukong | Minimum Framerate – Overall Average | 65.0 | ||
| Metro Exodus | Average Framerate – Runs | 123.3 | 123.0 | 122.7 |
| Metro Exodus | Minimum Framerate – Runs | 71.9 | 70.9 | 71.1 |
| Metro Exodus | Average Framerate – Overall Average | 123.0 | ||
| Metro Exodus | Minimum Framerate – Overall Average | 71.3 | ||
| Final Fantasy XIV | Average Framerate – Runs | 217.6 | 221.7 | 220.0 |
| Final Fantasy XIV | Minimum Framerate – Runs | 76.0 | 75.0 | 76.0 |
| Final Fantasy XIV | Average Framerate – Overall Average | 219.8 | ||
| Final Fantasy XIV | Minimum Framerate – Overall Average | 75.7 | ||
| Total War: Warhammer 3 | Average Framerate – Runs | 86.3 | 85.7 | 86.0 |
| Total War: Warhammer 3 | Minimum Framerate – Runs | 57.0 | 56.0 | 57.0 |
| Total War: Warhammer 3 | Average Framerate – Overall Average | 86.0 | ||
| Total War: Warhammer 3 | Minimum Framerate – Overall Average | 56.7 | ||
| 2560 x 1440 – GAME TESTING | ||||
| Cyberpunk 2077 | Average Framerate – Runs | 63.5 | 63.5 | 63.5 |
| Cyberpunk 2077 | Minimum Framerate – Runs | 56.4 | 56.2 | 56.3 |
| Cyberpunk 2077 | Average Framerate – Overall Average | 63.5 | ||
| Cyberpunk 2077 | Minimum Framerate – Overall Average | 56.3 | ||
| Black Myth Wukong | Average Framerate – Runs | 66.0 | 66.0 | 66.0 |
| Black Myth Wukong | Minimum Framerate – Runs | 55.0 | 53.0 | 52.0 |
| Black Myth Wukong | Average Framerate – Overall Average | 66.0 | ||
| Black Myth Wukong | Minimum Framerate – Overall Average | 53.3 | ||
| Metro Exodus | Average Framerate – Runs | 95.8 | 95.6 | 95.6 |
| Metro Exodus | Minimum Framerate – Runs | 61.6 | 63.0 | 62.7 |
| Metro Exodus | Average Framerate – Overall Average | 95.7 | ||
| Metro Exodus | Minimum Framerate – Overall Average | 62.4 | ||
| Final Fantasy XIV | Average Framerate – Runs | 185.1 | 187.2 | 186.7 |
| Final Fantasy XIV | Minimum Framerate – Runs | 73.0 | 74.0 | 74.0 |
| Final Fantasy XIV | Average Framerate – Overall Average | 186.3 | ||
| Final Fantasy XIV | Minimum Framerate – Overall Average | 73.7 | ||
| Total War: Warhammer 3 | Average Framerate – Runs | 81.8 | 81.5 | 81.7 |
| Total War: Warhammer 3 | Minimum Framerate – Runs | 54.0 | 55.0 | 54.0 |
| Total War: Warhammer 3 | Average Framerate – Overall Average | 81.7 | ||
| Total War: Warhammer 3 | Minimum Framerate – Overall Average | 54.3 | ||
| 3840 x 2160 – GAME TESTING | ||||
| Cyberpunk 2077 | Average Framerate – Runs | 34.7 | 34.7 | 34.6 |
| Cyberpunk 2077 | Minimum Framerate – Runs | 30.0 | 30.2 | 30.1 |
| Cyberpunk 2077 | Average Framerate – Overall Average | 34.7 | ||
| Cyberpunk 2077 | Minimum Framerate – Overall Average | 30.1 | ||
| Black Myth Wukong | Average Framerate – Runs | 40.0 | 40.0 | 40.0 |
| Black Myth Wukong | Minimum Framerate – Runs | 32.0 | 32.0 | 32.0 |
| Black Myth Wukong | Average Framerate – Overall Average | 40.0 | ||
| Black Myth Wukong | Minimum Framerate – Overall Average | 32.0 | ||
| Metro Exodus | Average Framerate – Runs | 58.2 | 57.9 | 58.1 |
| Metro Exodus | Minimum Framerate – Runs | 40.4 | 38.7 | 39.9 |
| Metro Exodus | Average Framerate – Overall Average | 58.0 | ||
| Metro Exodus | Minimum Framerate – Overall Average | 39.7 | ||
| Final Fantasy XIV | Average Framerate – Runs | 110.8 | 110.9 | 110.7 |
| Final Fantasy XIV | Minimum Framerate – Runs | 72.0 | 72.0 | 72.0 |
| Final Fantasy XIV | Average Framerate – Overall Average | 110.8 | ||
| Final Fantasy XIV | Minimum Framerate – Overall Average | 72.0 | ||
| Total War: Warhammer 3 | Average Framerate – Runs | 64.3 | 64.3 | 64.3 |
| Total War: Warhammer 3 | Minimum Framerate – Runs | 44.0 | 44.0 | 43.0 |
| Total War: Warhammer 3 | Average Framerate – Overall Average | 64.3 | ||
| Total War: Warhammer 3 | Minimum Framerate – Overall Average | 43.7 | ||
Ray Tracking, DLSS and AI Performance
- Improved transformer model
- Impressive performance boost with DLSS
First up, let me just say that the new transformer model Nvidia is utilising for its DLSS and frame-gen escapades is far better than its convolutional predecessor.
It’s cleaner, crisper, and latency has been dramatically improved, at least in my experience with it. MFG equally really did stretch its legs quite well here, and in my brief time dabbling with it in Cyberpunk, it produced average frames at nearly 200 fps in 4x mode.
It’s a bit frustrating really, as this is tech that is locked behind the RTX 50 series, despite the 40 and 30 series theoretically having the capacity and hardware to utilise it as well. They still can get that 2x mode, however.
Still in testing, the RTX 5080 did see an impressive performance boost utilising DLSS and frame gen in Cyberpunk and Black Myth. I kept the MFG at just 2x frames for Cyberpunk and average frames shot from 60fps with just DLSS to 105.3 fps with frame gen added on top.
| GAME | DLSS + FRAME GEN TESTS @ 4K | – FPS – | ||
| Cyberpunk 2077 | Average Framerate – Runs – DLSS ONLY | 59.8 | 59.9 | 59.9 |
| Cyberpunk 2077 | Minimum Framerate – Runs – DLSS ONLY | 53.1 | 53.4 | 53.3 |
| Cyberpunk 2077 | Average Framerate – Overall Average | 59.9 | ||
| Cyberpunk 2077 | Minimum Framerate – Overall Average | 53.3 | ||
| Cyberpunk 2077 | Average Framerate – Runs – DLSS + FG | 105.5 | 105.0 | 105.3 |
| Cyberpunk 2077 | Minimum Framerate – Runs – DLSS + FG | 95.6 | 93.6 | 94.4 |
| Cyberpunk 2077 | Average Framerate – Overall Average | 105.3 | ||
| Cyberpunk 2077 | Minimum Framerate – Overall Average | 94.5 | ||
| Black Myth | Average Framerate – Runs – DLSS ONLY | 40.0 | 40.0 | 40.0 |
| Black Myth | Minimum Framerate – Runs – DLSS ONLY | 33.0 | 33.0 | 32.0 |
| Black Myth | Average Framerate – Overall Average | 40.0 | ||
| Black Myth | Minimum Framerate – Overall Average | 32.7 | ||
| Black Myth | Average Framerate – Runs – DLSS + FG | 65.0 | 65.0 | 66.0 |
| Black Myth | Minimum Framerate – Runs – DLSS + FG | 56.0 | 55.0 | 56.0 |
| Black Myth | Average Framerate – Overall Average | 65.3 | ||
| Black Myth | Minimum Framerate – Overall Average | 55.7 | ||
In UL’s Procyon AI benchmark suite, results were quite promising. Nvidia’s TensorRT API delivered a score of 4274 in Computer Vision, versus the RTX 4080 Super’s 3728, although again using Windows ML, that deficit did once again drop.
Image generation equally saw significant improvements utilising Tensor over last gen, with ONNX only leading to a 10% increase in overall performance.
Power Consumption and Temperature
So, good news first. Nvidia has clearly proven with its Founders Edition that you can get away with just utilising a twin-slot cooler design for its RTX 5080 cards.
In my testing, max temps topped out at 76.0 C throughout the entire benchmarking process (ambient room temperature was 28.0 C). That’s such a big deal in today’s age of concrete-block-shaped graphics cards; it’s nice that we’ve got coolers available that really shrink them back down to a reasonable size without compromising on temperatures as well.
| POWER & TEMPS | |||
| Ambient Room Temperature | 28.0 °C | ||
| Max GPU Temp During Testing | 76.0 °C | ||
| Max Power Draw During Testing | 615.8 W | ||
Max power draw during testing was somewhat higher than on the previous generation, with a bump of around 30W or so under load, but that’s to be expected as TGP has gone up by 40W officially too, no doubt in part due to the increased clock speeds Nvidia is leveraging here.
Buy GeForce RTX 5080 Graphics Cards: Amazon / Currys / Laptops Direct / Overclockers / Scan
Should you buy it?
You want epic 4K performance, and are a big fan of DLSS and MFG
There’s no denying Nvidia’s GeForce RTX 5080 delivers some seriously impressive frame rates at 4K. All you need to do is grab a solid PSU and one of the best monitors you can find, and you’ll have a seriously good time with any title you choose to throw at it. With DLSS and MFG there if you need them, it effectively makes 4K a reality without breaking the bank.
You’re still on the 40 series, or are looking for a good deal
If you’re still on the last gen of Nvidia cards, unless you’re on an RTX 4060, it’s probably not worth jumping over to the RTX 5080 just yet.
Although it does deliver solid performance, that middling 12.9% improvement on last gen is hard to see when you’re actually gaming and only exists at 4K. If you can find one in stock for its retail price, even then, it’s not the best value for money card either (that title sits thoroughly with Intel’s Arc B580).
Final Thoughts
Nvidia’s GeForce RTX 5080 is a bit of an anomaly. A black cat in the world of graphics card launches. Its release has been marred with controversy and stock issues, not helped by global events and politics, and pricing remains often high.
In isolation (and this is certainly applicable if you’re on the 30 or 20 series from Nvidia), the RTX 5080 is a easily one of the best graphics cards, capable of delivering some exceptional 4K performance and backed up by an incredible suite of AI hardware to deliver some of the best high-res gaming experiences money can buy (outside of the RTX 5090).
How we test
Every GPU I get in for review is installed in its own dedicated test setup, complete with hardware that should ensure bottlenecks remain at a minimum.
Each card is then monitored via power meters and software to keep a check on power draw and temperatures while the benchmarking process is underway.
We utilize a total of 67 data points per card, testing across all manner of games and titles, plus synthetic benchmarks, at every mainstream resolution, and also test AI functionality as well.
- Tested with a dedicated setup
- Power monitored
- 67 data points tested
