Nvidia RTX 4090 vs RTX 4080: Battle of the new generation (2024)

Nvidia’s RTX 4090 and RTX 4080 are two staggeringly powerful graphics cards. Based on the new-generation Ada Lovelace architecture, they pack more CUDA cores, higher clock speeds, and more and faster memory onto boards that demand a lot of power and a lot of cooling. But the performance is worth it.

When it comes to 4K gaming, these cards are set to be some of the most powerful out there over the next year, and though there is new competition from AMD’s RX 7000 series of GPUs, Nvidia seems poised to retain the top slot in gaming performance. At least for now.

So if you’re looking for flagship gaming performance in the near future, Nvidia’s RTX 4090 and RTX 4080 are well worth considering. But which should you pick, and how do these cards compare in the real world?

Let’s take a look at their features and performance, and see if that makes them worth the high price of entry.

RTX 4000 – An Overview

The Nvidia RTX 4000 range of graphics cards is a brand new generation of GPUs built on Nvidia’s Ada Lovelace architecture, using the new TSMC 4N (5nm) process node. Although at the time of writing it’s still in its fledgling phases, with only two GPUs officially announced (one other announced, and then cancelled), we’re already seeing some exciting performance out of these cards and it shows a lot of promise and potential for the RTX 40-series as a whole.

The Ada Lovelace architecture is designed to increase CUDA core counts dramatically over the previous RTX 30-series, and allows for far faster clock speeds too. It also introduces new generations of RT cores for ray tracing, and Tensor cores, for deep learning super sampling (DLSS).

There’s been some clever tweaks to the rendering pipeline for ray tracing to further improve performance, and the new generation Tensor cores provide support for the latest DLSS 3, with its clever frame generation technology. That will remain a feature of RTX 4000 cards, for now.

This generation also removes NVLink support from even the top cards, so SLI gaming is now fully dead and buried, even if support was barely there on the last two generations of GPUs.

While there was a lot of talk about the potential for high power draw with the new generation cards when they launched, and they are rather power-hungry, Nvidia has changed what its TDP estimates actually mean with this generation.

Where previously it was more of a power target for graphics cards running at peak performance, it’s now more of a power limit. That should mean that the top cards don’t pull quite as much power as their TDP suggests in everyday gaming, nor spike over it like the RTX 30-series was wont to do.

New Generation, New Power

That said, these new cards do make use of the new 16pin power connector, sometimes referred to as the 12VHPWR connector, which can deliver far more power than traditional six and eight pin power supply cables. This is officially supported by some new power supplies, but for anyone with an older PSU there are adapters included with each card that let you connect multiple eight pin PCIexpress power cables to run the new cards.

There has been some controversy over adapters that melted during use, but these are extremely limited instances and in most cases were the result of user error. With that in mind, however, it’s a good idea when attaching the adapter, you ensure it’s plugged in fully and without any kinks in the wiring.

Image source: CCL

Nvidia RTX 4090 Vs 4080: Specifications

The specifications table is a great starting point for comparing two cards from the same generation. While it has limited scope in comparing cards between generations because of their different designs, when you have two cards based on the same process and a similar GPU, a specifications table can allow you to draw some strong conclusions about the cards’ potential.

Note: The above are specifications for the Nvidia Founders Edition versions of these GPUs. Cards from board partners will have different core and memory speeds, and possibly TDPs, too.

Spec Analysis

The differences between the RTX 4090 and 4080 are stark right from the get go.

CUDA core counts between them are almost incomparable, with the RTX 4090 sporting over 68% more CUDA cores. That should give it an enormous performance advantage in most games, especially since the clock speeds between the two cards are comparable – and actually slightly higher on the 4090, which was not the way previous generations of cards were specced out.

The RT and Tensor core difference is dramatic, too. While they both use the new, third generation RT cores and fourth generation Tensor cores, the 4090 will be much more performative. That would suggest that even if there is some measure of competition between the cards in games that use general rasterization, the 4090 should be head and shoulders above just about anything else when it comes to games with ray tracing.

Memory is another huge differentiator between the two, with the 4090 featuring 8GB more than its lower-tier counterpart. Both use GDDR6X memory, an Nvidia exclusive in its partnership with Micron, but the RTX 4080’s memory is actually slightly faster at 22.4 Gbps, rather than the 21.2 Gbps of the RTX 4090.

However, with a much wider memory bus on the 4090, there’s no wonder that the overall bandwidth is far higher on the flagship card. It demands extra power in turn, with the RTX 4090 getting close to its power limit according to most reviewers. The RTX 4080, on the other hand, reportedly regularly hits around 300W in gaming, so it is vastly more efficient, even if it’s unlikely to be able to keep up in performance.

Real World Performance of the RTX 4090 and 4080

Image source: Nvidia

The RTX 4090 and RTX 4080 are both stupendously powerful graphics cards that blow away everything from the previous generations at just about any resolution. But how do they compare against one another?

As you might expect from the lopsided specifications, the RTX 4090 is by far the more powerful card, delivering greater frame rates at all resolutions and detail settings than the RTX 4080, and especially so when ray tracing is enabled. But that’s not the whole story.

In synthetic benchmarks like 3DMark Time Spy, which focuses on traditional rasterization, the RTX 4090 can break through 31,000 points at stock and will no doubt be breaking records for the foreseeable future as overclockers push the card to its limit.

In comparison, an RTX 4080 paired with a decent processor can expect to get around 25,000, which is well above what the last generation could achieve (where even breaching 20,000 was rare). That’s a big jump for the 4080, but the RTX 4090 is in a whole other league.

The difference is even more pronounced in 3DMark Port Royal, which also looks at ray tracing performance. Here the RTX 4080 again puts in a valiant effort, almost reaching 18,000 points in some test benches. That’s around 50% more points than the RTX 3090 and 3080 Ti, so represents a big uplift in intergenerational ray tracing rendering performance.

But again the RTX 4090 is just so much more, with some reviewers hitting close to 26,000 points.

Gaming Benchmarks

These synthetic benchmarks are all well and good, but what about real games? Here, the results are similar, with the RTX 4090 still leaps and bounds ahead of the RTX 4080, but neither holds quite such stark leads over their predecessor.

Cyberpunk 2077 is still one of the most demanding games out there today and it really puts even the RTX 4090 through its paces. At 4K with all settings at Ultra, the RTX 4080 can almost hit 60 FPS average, but not quite. That gives it a 10-20% advantage over last-generation standouts like the RTX 3090 and 3080 Ti, making for a noticeable, if not revolutionary upgrade.

The RTX 4090 is that much faster again, hitting close to 80 FPS on average. That’s not super high frame rates, but it’s a big jump over the last-generation and cements once again that the RTX 4090 is the much more powerful card.

That’s even more true with ray tracing enabled, where with all settings maxed out (without DLSS) the RTX 4090 is the only card that can hit over 30 FPS average, managing around 40 FPS where even the best of the last generation barely breaks 20 FPS.

They’re Even Faster In Team Red Games

The same goes for Assassin’s Creed Valhalla, where the RTX 4080 is about 35% faster than its predecessors, but the RTX 4090 is that much faster again, able to hit over 120 FPS average at 4K with all the settings at Ultra High.

It is notable that in Valhalla, and in other games that do well on AMD cards (like Forza Horizon 5) that the RX 6950 XT and 6900 XT are within 10% of the RTX 4080, showing that it’s worth considering what games you’re going to play before making your purchasing decisions.

Results are a little closer between the two new Nvidia cards at 1440p, with the RTX 4090 roughly 20% faster than the RTX 4080 in that case. Considering the higher-end card will tend to get well over 150 FPS in any game at that resolution, though, it’s largely overkill.

Unless you’re an extreme esports gamer looking for every advantage, the new Nvidia flagship is very much a 4K GPU. You’d be spinning your wheels at anything else with such a powerful card.

DLSS 3 Frame Insertion, Is It Worth It?

Image source: Nvidia

Nvidia has been developing its deep learning super sampling upscaling algorithm for years and it’s gotten much better with each successive generation. Where DLSS 2 introduced much more nuanced upscaling with fewer artefacts, DLSS 3 is focused on a new feature: frame generation.

DLSS 3 can actually make up new frames based on the data from the game’s motion vectors and previous frames, allowing it to improve frame rates dramatically with only a very slight hit to frame times.

The result is enormous boosts in performance, like boosting the RTX 4080 from less than 30 FPS in Cyberpunk 4K with ray tracing, to over 120 FPS. Or improving the RTX 4090’s performance in a maxed-out Plague Tale: Requiem by over 100%.

The only problem with DLSS 3, at least in this early stage, is that its image quality isn’t stellar.

If you isolate the generated frames, they can often exhibit really funky visual artefacts, which in slower-paced games can be quite jarring. It’s especially problematic in games with heavy HUD elements, which aren’t always rendered in the same way as the background, and that’s quite confusing for the AI algorithm.

Is DLSS 3 a feature you’re likely to use? Maybe not straight away. If you’re spending over £1,000 on a GPU, you probably don’t want to compromise on image quality.

However, for games in the future where frame rates are far more limited or when these cards start to age and struggle a little, it’s a great feature to have and shows an exciting future for lower-end cards that may not be able to handle high end 4K gaming without it.

Conclusion

The Nvidia RTX 4090 and RTX 4080 are, until AMD’s next-generation cards launch (and likely beyond that) the two most powerful graphics cards money can buy. They showcase big leaps in performance over their predecessors and are arguably the best way to play many of the most demanding games.

However, when it comes to picking one over the other, you have a clear choice: The RTX 4090 arguably represents the best value in terms of performance. Frames per pound, the RTX 4090 is actually better value than the RTX 4080 at their current prices.

That said, do you really need an RTX 4090?

It’s so ungodly powerful that there are almost no games that can really push it and it is a lot more expensive. There are certainly no games that need such a powerful card to run and even run well. If money is no object and you want the absolute best, it’s an easy buy, but for everyone else, the RTX 4080 (and AMD’s upcoming cards) are well worth considering instead.