Nvidia has introduced their RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070 at CES 2025

For PC gamers, the debut of Nvidia’s RTX 50 series is likely to be the standout moment of CES 2025. Today, Nvidia has unveiled four new desktop graphics cards: the RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070.

With these new GPUs, Nvidia has implemented its latest Blackwell GPU architecture. This architecture features new hardware advancements, improved integration for tensor cores, and optimizations to technologies such as Shader Execution Reordering. The new Tensor cores in Blackwell now support FP4. Furthermore, Nvidia’s newly designed RT cores can handle twice as many ray-triangle intersections compared to previous models. Additionally, Blackwell now supports DisplayPort 2.1b with full bandwidth, allowing compatibility with 8K 165Hz displays.

With the RTX 50 series and Blackwell GPU architecture, Nvidia has released a new iteration of DLSS. DLSS 4 has been launched, replacing Nvidia’s Convolutional Neural Networks (CNNs) with a cutting-edge transformer model. This upgrade enhances the quality across all DLSS versions, including DLAA, Super Resolution, Ray Reconstruction, and Frame Generation.

Exclusively for the RTX 50 series is the capability of DLSS 4 Multi-Frame Generation. This feature allows for three AI-generated frames to be rendered for each real frame, significantly improving gaming framerates and enhancing the utility of ultra-high refresh rate gaming monitors.

Below, we can observe the effects of DLSS 4 and how it enhances the performance of Nvidia’s premium GPU lineup. With DLSS 4 activated, Nvidia’s RTX 5090 achieves double the performance compared to Nvidia’s RTX 4090. This is with full ray tracing enabled in Cyberpunk 2077.

With the new RTX 5090, Nvidia claims up to a 2x performance improvement over the RTX 4090. This is contingent on games implementing DLSS 4 and full ray tracing. In other titles, performance enhancements will be less dramatic. The RTX 5090 includes 32GB of GDDR6 memory, a total memory bandwidth of 1792 GBps, and features 21,760 CUDA cores. It is equipped with 680 5th generation Tensor Cores and 170 4th generation RT cores. Nvidia’s RTX 5090 will be available starting January 30th, with a retail price beginning at $1,999.

Similar to the RTX 5090, Nvidia promises a performance improvement of up to 2x with its RTX 5080 over its predecessor, the RTX 4080. Again, this is applicable in selected games with full ray tracing and DLSS 4. The RTX 5080 boasts 16GB of GDDR7 memory and provides users with 34% more bandwidth than an RTX 4080.  Nvidia’s RTX 5080 will also be available starting January 30th, with prices beginning at $999.

Nvidia guarantees a 2x performance increase with the RTX 5070 Ti compared to its previous model. This performance boost applies only to selected titles featuring full ray tracing and DLSS 4. The card will offer 16GB of GDDR7 memory, just like the RTX 5080. The suggested retail price for Nvidia’s RTX 5070 Ti will be $749.

Lastly, we have Nvidia’s RTX 5070, which Nvidia asserts to be 2x faster than its predecessor. Yes, this is again in selected games with DLSS 4 and full ray tracing enabled. This GPU will be equipped with 12GB of GDDR7 memory and a total bandwidth of 672 GBps. The MSRP for Nvidia’s RTX 5070 will be $549.

That concludes the keynote. The highlight was undoubtedly the announcement of the RTX 50-series, though there’s still much we don’t know. More details will be revealed in the coming days, and the full RTX 50-series launch is expected to kick off before the month ends.

Until then, remember to use our comparison tool to find the best deal for your next favorite game.