The RTX 5090 has finally been announced a release date confirmed, thanks to the impressive (and somewhat epic) Nvidia CES 2025 keynote presentation. As predicted, it looks like it’s going to wipe the floor with the RTX 4080, boasting arguably overkill specs at a whopping price point. While we haven’t yet had the chance to make the RTX 5090 run the gauntlet in our testing lab, as we did with the RTX 4080, we can use the specs to give us an idea of just how far apart these two cards are.
We will, of course, be putting the RTX 5090 through its paces and updating the article with real-world performance, so you can get a better idea of how they stack up against each other – stay tuned, as we’re not too far off its release date.
Spec comparison
Starting off with the specification difference between the two GPUs, we will compare how their basis has changed. Remember that being of different generations, they won’t be directly comparable, especially in architecture. While also being in different performance tiers in their lineups, there’s quite a discrepancy between them, but we see what that means for them.
| Specifications | RTX 5090 | RTX 4080 |
|---|---|---|
| CUDA Cores | 21,760 | 9,728 |
| Base clock speed | 2.01 GHz | 2.21GHz |
| Boost clock speed | 2.41 GHz | 2.51GHz |
| Memory | 32GB GDDR7 | 16GB GDDR6X |
| Memory interface | 512-bit | 256-bit |
| Bandwidth | 1.79 TB/s | 716.8GB/s |
| TGP | 575W | 320W |
CUDA cores
CUDA cores are the brain behind Nvidia’s graphics cards and the computing platform that makes them tick. So the more, the better, and as the top card, it’s unsurprising that the RTX 5090 has an incredible amount and is in a newer architecture, Blackwell. This means it uses a newer and better TSMC process to create its die, with a 4nm compared to 5nm in Ada on the RTX 4080.
In terms of numbers, the RTX 4080 has just 9,728 cores compared to the 5090’s 21,760. The new card offers over double the number of cores on board and would boost performance. Nvidia claims to have doubled the performance over the RTX 4090, so it likely would have more than doubled the performance over the RTX 4080, but it’s best to wait to see how benchmarks compare the two.
Nvidia also confirmed that the new DLSS 4 won’t be backward compatible, leaving the RTX 40 series with its own DLSS 3 and frame generation. As it pushes AI even more in its consumer cards, this should allow more neural processing and push performance even higher over the previous generation.
Clock speeds
The clock speeds are how quickly the CUDA cores process information and get those frames out to your screen. Here, it’s, in fact, the lower card that takes the lead, as the RTX 4080 has a better base and boost clock against the RTX 5090. Although not that much of a difference, that will bring the performance closer together, but it’s certainly an interesting spec to see.
That is also the factory clock, and without a GPU overclock, it can vary by users and custom models of the cards anyway. But by default, the 4080 has a 200Mhz higher base clock and a 100Mhz boost clock more than the 5090. If that were on the same card as an overclock, it would result in maybe a couple of FPS here and there, so it’s not a great ordeal.
Memory
Nvidia has upgraded its VRAM onboard the RTX 50 series specs. Opting for the next generation of memory means plenty more speed to get those textures and data loaded and buffered to help the graphics card perform. With the move to GDDR7 over GDDR6X, there is an excellent shift in both capacity and bandwidth in the 5090.
As it goes for a whopping 32GB onboard and a vast 512MB interface, it provides an incredible 1.79TB/s of data bandwidth for all the graphics buffering and loading into the VRAM and GPU. That towers over the RTX 4080s offering, which has just 16GB of GDDR6X memory, totalling a bandwidth of 716.8GB/s. With less than half of what the 5090 has onboard, it pales in comparison in terms of memory.
Power
Now, with such a behemoth of specs, with a shedload of memory, CUDA cores, and high clock rates, it’s no surprise the 5090 will drain your power quickly. As Nvidia sheds light on the expected total graphics power of the GPU, it has come to a massive 575W total and requires you to invest in a 1000W power supply at a minimum.
The RTX 4080 comes at just 320W and a 750W PSU requirement. It is something a bit more tame to house in your gaming PC and won’t be as draining on your electricity bills. It also shouldn’t produce as much heat with the lower power draw, even if the efficiency could be lower in the older card.
Price comparison
As for price, there’s quite a vast distance between the two. Well, less than if it were the 4080 Super, but there is a $800 difference between the two MSRP of the graphics cards. The RTX 5090 starting price is $1,999 compared to the RTX 4080’s launch price of $1,199, making it quite the stark difference at the top of the ladder.
This can be quite the significant deciding factor for your budget; even if they are already some of the top picks, there’s quite a big difference in costs.
Final word
There is quite a stark difference between the two graphics cards, which is unsurprising. Considering the RTX 5090 is the flagship, it has the full force of Blackwell and supporting specs behind it to create a top-notch GPU. But that comes at quite the cost, reaching the $2000 mark just for one PC component.
On the other hand, the RTX 4080 is a great gaming card for a lot less, as the Ada GPU still provides a whole load of performance without splashing out the extra for the new ones. You can also consider the RTX 4080 Super, which provides better value for money as it upped the CUDA cores for a lower price than the original 4080.
