Almost exactly a year after Nvidia realised the full potential of RTX 4080 with a Super card, here are we are with RTX 5080 on the scene. Remarkably similar to RTX 4080 Super on first glance, the devil is in the details. Nvidia swaps out Ada Lovelace for Blackwell and adds in more robust ray tracing, faster memory, hardware support for a new AI format, yet keeps price to the same $999 for Founders Edition and stock-clocked partner models. Colour me intrigued.
In this guide I’ll let you know whether the sum of changes means it’s worth buying over and above RTX 4080 Super and how its secret weapon is a promise of things to come.
But I can’t do all that unless I roll out the fabled Club386 Table of Doom™.
| RTX 5080 | RTX 4080 Super | Ratio | |
|---|---|---|---|
| Released | Jan 2025 | Jan 2024 | – |
| Codename | Blackwell | Ada Lovelace | – |
| GPU | GB203 | AD103 | – |
| Process | TSMC 4N | TSMC 4N | – |
| Transistors bn | 45.6 | 45.9 | 0.99 |
| Die size mm² | 378 | 378.6 | 1.00 |
| CUDA cores | 10,752 | 10,240 | 1.05 |
| Base clock MHz | 2,295 | 2,205 | 1.04 |
| Boost clock MHz | 2,617 | 2,550 | 1.03 |
| FP32 Boost TFLOPS | 56.3 | 52.2 | 1.08 |
| SM count | 84 | 80 | 1.05 |
| RT cores | 84 (4th Gen) | 80 (3rd Gen) | 1.05 |
| RT TFLOPS | 170.6 | 121 | 1.41 |
| Tensor cores | 336 (5th Gen) | 320 (4th Gen) | 1.05 |
| FP16 Acc TFLOPS | 450.2 | 418 | 1.08 |
| FP4 TOPS | 1,801 | 836 | 2.15 |
| ROPS | 112 | 112 | 1.00 |
| Memory GB | 16 | 16 | 1.00 |
| Memory type | GDDR7 | GDDR6X | – |
| Mem. clock Gb/s | 30 | 23 | 1.30 |
| Mem. interface bits | 256 | 256 | 1.00 |
| Mem. bandwidth GB/s | 960 | 737 | 1.30 |
| Board power watts | 360 | 320 | 1.13 |
| Launch MSRP $ | 999 | 999 | 1.00 |
Before you do anything else, take a look at the ratio column on the right-hand side. It defines how, on paper, these two heavyweights compare against each other. Notice startling similarity until you get to RT TFLOPS, AI TOPS, and memory clock? This is entirely deliberate on Nvidia’s behalf.
Both GPUs are based on ostensibly the same manufacturing process and house near-identical die sizes and transistor counts. This fact alone tells me Nvidia is more than happy with its existing $999 proposition, so all that’s needed is a better architecture and additional smarts.
RTX 5080’s core count and frequency is what gives it the leg-up in terms of FP32 throughput that’s closely tied with gaming rasterisation performance. Ray tracing sees a larger boost, mind you, as engineers have reworked the core to offer more throughput. AI inferencing also receives a shot in the arm with hardware support for the FP4 format, and should your workload use it, expect over twice RTX 4080 Super performance. I’m glad to see memory bandwidth gets a makeover, too, with GDDR7 running at a blistering 30Gb/s.
In terms of straight-up architecture, then, Nvidia’s kept things mostly the same, albeit with useful increases in areas which it thinks will benefit most from game code of today and tomorrow. RTX 5080 may not look all that impressive against RTX 4080 Super today, but that may not be the case in three years’ time.
Nvidia is also aware that throwing precious, expensive silicon at a GPU advancement problem no longer makes financial sense. Innovation absolutely needs to occur outside of this area. RTX 50 Series’ answer is a frame rate-boosting technique known as multi frame generation (MFG), which builds on standard frame generation launched with RTX 40 Series. I strongly encourage you to read all about it because it’s the most potent weapon in RTX 5080’s arsenal.
On the face of it, expect 10-15% gains in rasterisation performance, 20% for ray tracing, and more than doubling of frame rate when used in games fully leveraging MFG.
Let’s see how my predictions play out as I run through applications and games touching on each facet discussed above.
Performance
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
GeForce RTX 5080 Founders Edition faces up to GeForce RTX 4080 Series Founders Edition. They’re tested back-to-back, minimising potential software-related inconsistencies.
Application and AI
Relying on a heady mix of rasterisation and ray tracing, 3DMark Speed Way favours RTX 5080 to the tune of 20 per cent. Given what I know about architecture, that’s more than I was expecting, though it may well have to do with the new card simply running the ray tracing components of the test better.
Steel Nomad, meanwhile, surprises me more, as RTX 5080 pulls out a 25% lead. In the land of synthetic benchmarks, Blackwell reigns supreme. I’m putting this down to RTX 5080’s higher memory bandwidth than anything else.
Many of you use Blender for rendering purposes. Don’t go rushing out to buy the new GPU as it’s only 8% faster.
A similar story in Geekbench AI.
Procyon’s AI Text Generation benchmark takes the Llama 3.1 model and runs queries offline. The test evaluates how long the model takes to start a reasoned response and then calculates how quickly it’s able to add further context. Putting these together leads to a score. You’re looking at around 20% improvement.
Gaming
Assassin’s Creed Mirage doesn’t think too much of RTX 5080, mind you, as it’s only 9% quicker at UHD. Certainly not enough to justify a generational numbering upgrade.
I’m also seeing about the same in Final Fantasy XIV: Dawntrail. Part of the reason also has to do with Nvidia failing to increase the ROP count between generations, standing at 112, which is usually an easy lever to pull when demanding higher rasterisation.
Remember I said one of Blackwell’s key strengths is enhanced ray tracing? I see that in evidence as the new card puts almost 20% into the old one bearing a similar name.
Exact performance uplift form one generation to the next depends on how well the game taps into RTX 5080’s assortment of goodies. I feel as if higher memory bandwidth is the key driver here.
The same, actually, is also true in Tom Clancy’s Ranbow Six Extraction. Based on what I’ve seen thus far, I wouldn’t be swapping out RTX 4080 Super for RTX 5080.
Dialling up the hurt on the pair is Cyberpunk 2077 run with the Ray Tracing Overdrive preset that calls path tracing into play. With these settings, however, I’d only want to play at FHD, which is an odd thing to say about $999 cards.
DLSS
Nvidia’s DLSS upscaling technology helps almost double frame rate for both cards. Numbers aside, the pair feels experientially practically the same.
Recall the virtues of MFG? RTX 5080 doubles RTX 4080 Super’s frame rate at all resolutions. It sure looks good on paper but be mindful that an MFG-driven 4K 137fps doesn’t feel like a native 137fps. That’s because Nvidia’s technology uses the base 19fps as the foundation on which to build frame generation upon, so there’s always more latency than you’d think. I’d ideally like to see this base number at 40-60fps for it to work well.
| Game | 4080 Super to 5080 % uplift at 4K |
|---|---|
| Assassin’s Creed Mirage | 8.9 |
| Final Fantasy XIV Dawntrail | 10.9 |
| Forza Motorsport | 18.0 |
| Mount and Blade | 16.9 |
| Rainbow Six Extraction | 14.1 |
| Cyberpunk 2077 DLSS + FG | 114 |
Taking all the games at 4K and comparing results, I see between 8.9% and 18% improvement between generations in everything other than MFG. Is that enough? One can make a coherent argument either way, but as far as I’m concerned, I was expecting a tad more, to the tune of 20-25% before MFG.
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 3080
GeForce RTX 5080 vs. AMD Radeon RX 7900 XTX
Vitals
GB203 and AD103 are physical facsimiles of one another. No surprise to see near-identical power consumption.
Both Founders Edition coolers do a good job in keeping cards cool.
Don’t be too surprised by the above graph. If you recall, RTX 4080 Super uses a much larger cooler whose fans run more quietly.
Conclusion
Today’s $999 champion takes most of what’s good with last-generation RTX 4080 Super and adds enhanced ray tracing, FP4-format AI, and faster memory. Oh, did I forget to mention MFG?
I can see why some gamers may be disappointed with GeForce RTX 5080 if they predominantly play older games. Nvidia’s Blackwell advancements speak to the future, not the past, so you’re oftentimes left with single-digit gains over its immediate predecessor.
I’m actually okay with this because one cannot expect 30-50% improvements in every metric, every generation. I’d rather investment occur in novel techniques that have innate ability to fundamentally change your games-playing experience than simply balloon transistor count by laying down more of what’s made previous generations tick along. CUDA FP32, you’re no longer the favourite child.
Nvidia follows this to the letter with RTX 5080 and MFG. Available on 75 games and apps at launch through either native support or control panel override, I’m hopeful more of the big studios will take the time to implement in-game support. Wait and see.
Starting from scratch and building a premium PC, I’d absolutely buy GeForce RTX 5080 over 4080 Super, of course, but I wouldn’t necessarily upgrade from the year-old card.
GeForce RTX 5080
“Gear up for game-changing experiences with the Nvidia GeForce RTX 5080 and AI-powered DLSS 4.” Read our review.
GeForce RTX 4080 Super
“Enjoy phenomenal gaming and content creation performance with the Nvidia GeForce RTX 4080 Super.” Read our review.
