NVIDIA GeForce RTX 5080 FE review: the perfect balance between power and price?

You wait years for one, and then two come along at once. Yes, the NVIDIA GeForce RTX 5080 graphics card is here, and we've taken it for a test drive, too see how it will perform for power-hungry creatives.

While the flagship card, the GeForce RTX 5090 FE (FE stands for Founders Edition), naturally comes with All The Power, I was intrigued to see how close the 5080 FE could run it, to see if it would be a more affordable top choice of graphics cards for graphic artists and 3D pros. And you know what? It was a closer race than I thought.

NVIDIA GeForce RTX 5090 FE review: Key specifications

Design and build

The first thing I noticed as I unpacked the NVIDIA GeForce RTX 5080 FE from its fully recyclable packaging (which looked identical to the 5090 apart from the number at the top), was that it looks almost exactly the same as the 5090 card. The only noticeable difference is the model number etched into the topside (and that it's ever-so-slightly more prominent on the 5080, although that could just be my specific unit). A solid, weighty piece of hardware, slightly longer and narrower than the 40-series cards, it slots in very easily but leaves a little daylight in the back-panel notch on my tower. Thankfully, the only witnesses to that will be my wall and my overly curious cat.

The power socket being inserted at an angle saves space, as the cable juts out less from the side of the card than it otherwise would, but inserting and removing the cable is a little bit more finicky as a result. Hardly a deal-breaker though...

The GEFORCE RTX logo lights up too, and there's light protruding through from inside too, most prominent in the V-shaped corners near the centre as you look at it from above.

The back, just like the 5090, has four ports, three DisplayPort and one HDMI. You'll be entirely unsurprised to find the DPs support up to 4K 480Hz or 8K 165Hz with DSC each, while the HDMI supports up to 4K 480Hz or 8K 120Hz with DSC, Gaming VRR, and HDR.

Features

The NVIDIA 5080 comes with the same feature and software offering as the 5090, sporting the same Blackwell core architecture, which has stepped up both AI-enhanced output and efficiency, while providing less of an uptick in pure VRAM terms.

In fact, the 5080 comes with 16GB of the latter, the same number you'll find inside the 4080 version. So how is the new one better, I hear you howl in abject confusion at your favourite reviews editor?

Well, I went to NVIDIA's Editors Day at CES 2025, and as I learned there, the AI so many of us hate to see used to create pictures of people with varying numbers of fingers and other appendages or dodgy boomer memes on Facebook, has been employed here to provide a dramatic uptick in power efficiency.

For example, this is apparent in how it can shut down a majority of its cores more quickly during idle times during graphics rendering, which in turn means those cores are available to help with new tasks while the previous ones are still running, instead of those cores being spun up and doing nothing productive on said previous task.

This means that with those AI-enhanced features activated, we can expect to see up to 40% higher performance from the same amount of VRAM during many 3D tasks, (and like I pointed out already in the 5090 review, multi-tasking in video rendering means that time-consuming job has become a lot faster than that).

Yes, Mega Geometry, Neural rendering, DLSS 4 and a host of other tools will also be available on the new drivers for the 30- and 40-series of cards, but they don't have the added benefit of the Blackwell core architecture, if those not-so-marginal gains are important to you.

Tensors

NVIDIA's proprietary technology, CUDA cores, is central to the graphics performance of its GPUs. For many professionals in the 3D field, the number of CUDA cores on a card has been the primary factor in selecting a GPU.

Additionally, the increasingly important Tensor Cores play a crucial role. Both CUDA cores and Tensor Cores are processing units, and having more of them generally improves performance. While CUDA cores excel at general graphics and 3D rendering, they are less efficient for neural processing, AI, and deep learning tasks compared to Tensor Cores.

The GeForce 5080 Founders Edition (FE) features 1,801 Tensor Cores, which is over 30% more than the 1,321 found in the GeForce 4090. However, the 5080 FE has 10,752 CUDA cores, which is fewer than the 16,384 CUDA cores in the 4090.

Benchmark scoring

Performance

In all honesty, I expected a bigger dropoff in benchmark and real-world performance levels from the 5090 FE here. Unless you're working on the latest Pixar movie or a billion-dollar AAA game, the 5080 might just be enough for you for the next few years.

PugetBench for Creators testing gave me near-identical numbers in the Photoshop test to what the 5090 delivered. I asked NVIDIA about this and the answer was that "there appear to be some scaling issues with the pre-launch build which means the results might not be completely representative of performance difference between the cards." This might already have been sorted out by the time you receive your card.

I also used 4-2-2 camera video assets to run a concurrent render of 9 video streams, and again, the 5080 gave my 4070 Ti Super and its single render job a run for its money. I estimated about a 8.5x increase in speed here, which should vastly speed up work for many video editors.

When it comes to Procyon Flux testing, we finally see a marked difference between the models. The 5080 FE took more than twice as long to generate images using the memory-demanding FP4 test than the 5090 FE did, but it was still impressively fast in its own right.

There's also about a 15% reduction in performance from the 5090 to the 5080 in both the PugetBench DaVinci test and the Topaz Video AI upscaling test. Both figures still reside firmly in the 'staggering' category though.

Also, H.264 encoding saw no dropoff in the 5080, with the lower-spec card churning through 128.98 frames per second, all-but identical to the 5090's 129.04 frames per second.

And while the Geekbench 6 OpenCL testing numbers didn't match-up to the reality-bending figures churned out by the 5090, the 5080 FE produced a hugely respectable 272,788 points here, outscoring the 4090 in a test that's not yet been optimised for the AI-enhanced features of the Blackwell-equipped card.

3D processing is now drastically faster, thanks to innovations in the 50-series' feature offering. While software support is still being rolled out, it is clear that by the end of the year, nearly all major 3D and video software will have integrated the Blackwell suite of features. That means some of our standard benchmark tests are not fully compatible with the 50-series graphics cards. We will update this review with those tests as soon as we have the opportunity to run them.

Gaming is also truly spectacular here, especially when I can make use of features like DLSS 4. I ran Cyberpunk 2077, EA FC 24 and more games on this card, and the step up from 4070 Ti Super I had previously is not only noticeable, it's staggering. Go to one of our gaming-focused sister sites for a full breakdown of the card for gamers if you're interested to know more about that.

Also, extra point for releasing an impressively stable driver on the first go.

Who is it for?

At under a grand at launch, the NVIDIA GeForce RTX 5080 FE isn't exactly cheap, but while it's only half the price of the 5090 FE, it provides a heck of a lot more than half the performance.

With DLSS 4 support, neural rendering, RTX Mega Geometry and a host of other innovations on board, 3D modellers, animators and perhaps most importantly game developers will be looking for the most bang for their buck when it comes to the 50-series. And the 5080 might be that card.

Meanwhile, I'd actually recommend the 5080 over the 5090 for video editors regardless of the price benefit. Same for photo editors (although a 40-series card will remain more than enough power for most of you).

Buy it if

• you are a pro game dev, animator or 3D artist
• you need near-top-level performance but your budget isn't unlimited
• you want the best card for video editing

Don't buy it if

• a 40 (or 30) series card will do the job