Nvidia introduces new RTX 5090, 5080 and 5070 graphics cards

Nvidia and its CEO Jensen Huang gave a keynote at the start of CES 2025, during which the new graphics cards RTX 5090 (1999 $), RTX 5080 (999 $), RTX 5070 Ti (749 $) and also RTX 5070 (549 $) were introduced. We do not yet have European prices, but the RTX 5090 and RTX 5080 will be available on January 30th, while the RTX 5070 Ti and RTX 5070 will be available sometime in February.

The price of the RTX 5070 models is a pleasant surprise, as it is lower than when the RTX 4070 cards were released. Whether this will also be the case when cards from different manufacturers are released remains to be seen. However, there must always be a balance, which is why the RTX 5090 400 $ is more expensive than the retail price of the RTX 4090.

The RTX 50 Series graphics cards feature a new Founders Edition design with just two dual-flow fans, a 3D vapor chamber, and GDDR7 memory. All RTX 50 Series cards are PCIe Gen 5 and include DisplayPort 2.1b connectors for connecting displays up to 8K and 165 Hz.

Nvidia RTX 5090 – specifications

What's a bit surprising is the size of the new RTX 5090 Founders Edition. It will take up just two slots in your computer, which means it will be much easier to install in smaller computers compared to its predecessor, the RTX 4090, which was really huge by graphics card standards. The RTX 5090 has 32 GB of GDDR7 memory, 1792 GB/s of memory bandwidth, and a whopping 21,760 CUDA cores. It will have a whopping 3352 TOPS of performance for AI tasks.

Together with DLSS 4 technology and the Blackwell architecture, the RTX 5090 is said by Nvidia to be twice as fast as the RTX 4090. We can't wait for the testers who will be among the first to test this beast to confirm this. The consumption will understandably be higher. Nvidia says that the total graphics power of the RTX 5090 will be 575 watts, and the recommended power consumption is 1000 watts. This is 125 watts more than the RTX 4090 model, but we will have to wait and see what they have done in terms of efficiency or in which scenarios the card will use all the available power.

Nvidia demonstrated the power of the new card with the example of the game Cyberpunk 2077, which runs at 238 frames per second (fps) on a computer with an RTX 5090 with DLSS 4, compared to 106 fps on a computer with an RTX 4090 with DLSS 3.5. The game was run with ray tracing enabled on both GPUs.

Nvidia RTX 5080, RTX 5070 Ti, RTX 5070 – specifications

The RTX 5080 graphics card is also said to be twice as fast as the RTX 4080 and will feature 16GB of GDDR7 memory, 960GB/s memory bandwidth, and 10,752 CUDA cores. The Nvidia RTX 5080 will have a total graphics power of 360 watts, and users are recommended to purchase at least an 850-watt power supply. In terms of artificial intelligence functions, it will have about half the performance (1,801 TOPS) of its more powerful sibling.

Nvidia also introduced the RTX 5070 Ti and RTX 5070 cards. The RTX 5070 Ti has 16GB of GDDR7 memory, 896GB/s memory bandwidth, and 8960 CUDA cores. The regular RTX 5070 has 12GB of GDDR7 memory, 672GB/s memory bandwidth, and 6144 CUDA cores. The RTX 5070 Ti has a total power draw of 300W and requires a 750W power supply, while the RTX 5070 has a total power draw of 250W and requires a 650W power supply.

Nvidia says these cards will also be twice as fast as their immediate predecessors. CEO Jensen Huang even claimed on stage at CES that the RTX 5070 will deliver “RTX 4090 performance for $549,” but likely with a strong boost from DLSS 4 technology rather than raw performance.

Nvidia is also preparing its RTX 50 series for laptops, albeit with slightly reduced specifications. The RTX 5090 will debut for laptops with 24 GB of GDDR7 memory. The portable version of the RTX 5080 card will be available with 16 GB of GDDR7 memory, the RTX 5070 Ti with 12 GB of GDDR7 memory, and the RTX 5070 with just 8 GB of GDDR7 memory. The RTX 50 series laptops will be available from various PC manufacturers starting in March.

New shaders, more artificial intelligence

It’s not over yet. Huang and co also showed off Nvidia’s new RTX Blackwell cores and gave a demo of their rendering capabilities. The demo included new RTX Neural Materials, RTX Neutral Faces, text-to-animation, and DLSS 4. “The next generation of DLSS can do more than just render frames, it can predict the future,” Huang says. “We enabled AI with GeForce, and now AI is revolutionizing GeForce.”

Nvidia's new RTX neural shaders can be used to compress textures in games, while RTX neural faces aim to improve the quality of faces using generative AI. The next generation of DLSS includes Multi Frame Generation, which creates up to three additional frames per traditional frame and can increase rendering speeds by up to 8x compared to traditional rendering, according to Nvidia.

DLSS 4 also includes the use of real-time transformers to improve image quality, reduce glare, and add more detail to motion. The DLSS 4 upgrade will even work on existing RTX cards, but it's unclear if this includes just the latest-generation RTX 40 or perhaps the RTX 30 as well.

We've been waiting for them for two years, and now we've got them. From what we've seen, we can say that Nvidia has met the high expectations. We're all eagerly awaiting the first tests and comparisons with the previous generation and also the competition that AMD is preparing, which is more focused on the mid-range price range.