RTX 5090 vs RTX 4090 specs comparison – how do these flagship GPUs differ on paper?
Table of Contents
The RTX 5090 has finally launched, and it looks seriously impressive. Of course, the first thing we thought about was how the RTX 4090 would look in comparison, especially since the RTX 5070 is claimed to offer the same performance but for way less cash.
We’re planning on putting the RTX 5090 through its paces in our testing lab, much like what we did with our RTX 4090 review. But the review embargo is already up, and there are plenty of write-ups and videos on its performance already, so we’re going to give an overview of some of the key differences using what’s available now.
Here’s how the best 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
| Specifications | RTX 5090 | RTX 4090 |
| Cores | 21760 | 16384 |
| Base clock speed | 2.01 GHz | 2.23 GHz |
| Boost clock speed | 2.41 GHz | 2.52 GHz |
| Memory | 32GB GDDR7 | 24GB GDDR6X |
| Memory interface | 512-bit | 384-bit |
| Memory clock | 28 Gbps | 21 Gbps |
| Bandwidth | 1.79 TB/s | 1.01 TB/s |
| TDP | 575W | 450W |
| Power connector | 12V-2×6 | 12V-2×6 |
| Release date | Jan 2025 | Oct 2022 |
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 technology. This gives the 5090 the edge, and there are plenty of DLSS 4 games already planned to offer the 50 series, so much opportunity to thrive.
Other features include Reflex 2, which is the latest Nvidia technology designed with reducing latency in mind. This is a bit of a 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 are a competitive gamer, I’d say it’s another reason to go for the RTX 5090.
Sizing
One of the big talking points of the release is the much-improved sizing of the Founder’s Edition Nvidia showed off. Cutting down the sizing of the GPU even though the board power has increased. As the official specs show, the 575W card only takes up two slots with a length of 304mm and a width of 137mm. The 4090 FE has the same length and width but takes up 3-slots, even with a 450W power requirement.
Nvidia shows off the design and process of creating it in a video and how it managed to achieve that. Minimizing the actual PCB and allowing the fans to blow right through the whole card to cool it down. While the 4090 did cut the size of the PCB down to allow one fan to blow through, the 5090 expands on that and allows it to be one slot smaller. Also, allowing a better SFF compatibility with smaller builds.
Performance
Although we have yet to test the card ourselves, many reviews have gone live with the performance values. Specifically comparing how much of an uplift the 5090 vs 4090 provides. So, going through various benchmarks and reviews, we can see what kind of improvement it provides.
Looking through a few of them, you can expect to see a 20-50% increase over the 4090 in 4K rasterization. While in ray tracing, that is around 27-35%. Some consolidation of those results results in around a 12% performance gain and a 27% performance average at 1440p and 4k, respectively.
With citations, the CPU tends to be bottlenecking the GPU at those lower resolutions, where it cannot keep up with the power of the 5090. Most don’t even bother showcasing the 1080p performance for this reason, so you really have to have the best CPU for the 5090 for the job.
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 delivers 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!
-
ASUS ROG Strix RTX 4090 OC
- GPU: AD102
- CUDA cores: 16,384
- VRAM: 24GB GDDR6X
- Memory bus: 384-bit
- Bandwidth: 1,008 GB/sec
- Base clock: 2235 MHz
-
-
About the Author