RTX 5090 vs RTX 4090 specs comparison – how do these flagship GPUs differ on paper?

The RTX 5090 has finally been announced, and it looks seriously impressive. Of course, the first thing we thought was how the RTX 4090 was going to look in comparison, especially as the RTX 5070 is claimed offer the same performance, but for way less cash…

The RTX 5090 is its way over this month, and we’re planning on putting it through its paces in our testing lab, much like what we did with our RTX 4090 review. But the good news is that thanks to CES 2025, we finally have some official specs to work with, which can kick start our comparison between the two cards.

We’ll be updating this article with real-world performance once our review is live, but until then, here’s how the RTX 5090 and RTX 4090 differ on paper, and just how big the performance gulf looks to be.

RTX 5090 vs RTX 4090 specs comparison

Since both of these GPUs are flagships of their respective generations, the hardware is impressive, but the RTX 5090 has more decked-out hardware, especially the VRAM, memory type, and CUDA cores. 

Memory 

The first and most significant difference between these GPUs is that the 5090 has 32GB of GDDR7 VRAM, while the RTX 4090 has 24GB of GDDR6X VRAM. While the RTX 4090 is more than enough for any title currently available and can deliver excellent performance with DLSS 3, it seems like the 5090 took it to the next level. 

32GB GDDR7 VRAM is definitely overkill for gaming, but for productivity, the more, the merrier as tasks become more complex, and more data must be processed to reflect good performance and efficiency. That being said, the RTX 5090 will likely outperform the RTX 4090 by a good margin and take the victory in this domain. 

Clock speed

The RTX 4090 has higher base and clock speeds than the RTX 5090. It has a 2.23 GHz base clock speed and a 2.52 GHz boost clock speed, while the RTX 5090 has a base and boost clock speed of 2.01 and 2.41 GHz, respectively. 

This means that the 4090’s CUDA core is running faster and can process more data per clock cycle; however, even with slower clock speeds, the 5090 should have better performance due to the sheer difference in CUDA cores between the two GPUs. 

Cores

The RTX 4090 has 16,384 CUDA cores, while the RTX 5090 has 21,760 CUDA cores. So, the latter has 5376 more CUDA cores than the former, which means that the performance gap between these two cards can’t be bridged by software or overclocking. 

The RTX 5090 has significantly more cores pulling the weight, albeit at a slower clock speed, but more numbers mean more data processing in parallel, resulting in better and more efficient output. 

Features

The big focus of Nvidia’s CES keynote revolved around the usage of AI to further enhance performance. Notably, the RTX 5090 will have DLSS 4, which will assist in boosting FPS using its new Multi Frame Generation abilities.

Nvidia has an excellent video showing the differences between DLSS technologies that you can view below.

How does it all work? Well, without going too in depth, what it essentially does is use AI to ‘predict’ visual data that hasn’t yet been rendered. That can be anything from lighting to textures. If it’s predicting this in realtime, that means that you’ll see a drastic reduction in ghosting or any stuttering – both of which can take you out of the moment when you’re gaming.

This was shown off using a demo at the Nvidia CES 2025 keynote. I’d highly recommend taking a look as the results are simply jaw-dropping.

But how does this compare against the RTX 4090? Put simply, the 4090 has DLSS 3.5, which is a less advanced version of the tecnology. This gives the 5090 the edge.

Other features include Reflex 2, which is the latest Nvidia technology designed with reducing latency in mind. This is a bit of secret weapon when it comes to first person shooters in particular, as latency can mean the difference between victory and crushing defeat.

By reducing latency and using a feature called ‘Frame Wrap’, using The Finals as an example, Nvidia claims you can reduce latency by 75% overall.

Reflex 2 will debut on the RTX 50 series, and then be made available later on other RTX cards via an update. If you don’t fancy waiting or you’re a competitive gamer, I’d say it’s another reason to go for the RTX 5090.

RTX 5090 vs. RTX 4090 – price comparison

Depending on the hardware improvements of the RTX 5090, it isn’t surprising that it costs more than the RTX 4090. It has been announced with an MSRP of $1,999, while the RTX 4090 had a price tag of $1,599. 

However, third-party variants of the 4090 were priced anywhere from $2,000 to $2,500+. So, we could potentially see variants of the 5090 with a $3,000 price tag or even above that. 

Early impressions

There is no doubt that the RTX 5090 has much better hardware than the RTX 4090, and coupled with the new Multi-Frame Generation and DLSS 4, it should deliver the ultimate 4K experience even in the most graphically demanding titles. 

However, such performance comes at a cost, and few can afford such a high-end end or have a use for it. So, if you’ve got an RTX 4090 and can run all the games easily, there isn’t a need to upgrade to the RTX 5090 unless you’re a hardware enthusiast who wants the best of the best.

We’re looking forward to getting hands on with the new card and seeing what it brings to the table – stay tuned!