Nvidia has announced its latest generation of Geforce RTX graphics cards based on its Blackwell GPU architecture. The RTX 50 series will launch in late January, with four tiers of cards announced at the event. Nvidia CEO Jensen Huang also announced other additions to the company’s products during the presentation, such as Geforce NOW for Steam Deck, Nvidia Broadcast app updates and more.

You can check out our general breakdown of Nvidia’s CES announcements. However, if you’re just here to see how each new GPU performs, read on.

Nvidia Geforce RTX 50 GPUs: Specifications explained

The RTX 50 series is based on the Blackwell architecture and uses ninth-generation NVENC encoders, fifth-generation Tensor cores and fourth-generation ray tracing cores. Nvidia announced four graphics cards at the event: RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070.

Nvidia’s highest-end consumer graphics card is the RTX 5090, which is said to be twice as fast as the RTX 4090 in AI-powered gaming performance. With 3352 AI TOPS and 21760 CUDA cores, the RTX 5090 is an impressive GPU. It consumes a total of 575W of graphics power. That’s what packing 92 billion transistors performing over 3342 trillion AI operations per second will do.

The RTX 5070, the lowest tier card announced at CES, features 988 AI TOPS and 6144 CUDA cores. Compared to the previous generation RTX 4070, the 5070 has more than double the TOPS, resulting in significantly faster performance in AI applications.

“Blackwell, the engine of AI, has arrived for PC gamers, developers and creators,” said Jensen Huang, founder and CEO of NVIDIA. “By combining AI-driven neural rendering and ray tracing, Blackwell is the most significant computer graphics innovation since the introduction of programmable shading 25 years ago.”

Nvidia Geforce RTX 50 GPUs: Prices and availability

Here are the prices and release windows for the announced RTX graphics cards:

• Nvidia Geforce RTX 5070 – $549, February 2025 release
• Nvidia Geforce RTX 5070 Ti – $749, releasing in February 2025
• Nvidia Geforce RTX 5080 – $999, releasing January 30th
• Nvidia Geforce RTX 5090 – $1,999, releasing January 30th

The Nvidia Geforce RTX 50 series uses DLSS 4

The Geforce RTX 50 series uses a new model of DLSS upscaling and super resolution to deliver improved performance compared to the previous generation. This is because DLSS 4 uses what Nvidia calls Multi Frame Generation, which “generates up to three additional frames per traditionally rendered frame.”

Multi-frame generation via DLSS 4 will be available exclusively for the Blackwell graphics card family, although general improvements to DLSS capabilities will be available for existing cards. This is because the new DLSS features are based on a transformer-based model, which replaces the Convolutional Neural Network model. DLSS 4 is said to be “40% faster” and use “30% less VRAM”.

A total of 75 games have been confirmed to support DLSS 4 at launch, with more on the way. Nvidia also announced Reflex 2, an improved version of its latency-reducing technology. Reflex 2 uses “frame warp technology” to reduce latency by up to 75%. The feature appears for the first time in The finale And Valorant.

Geforce RTX 50 GPU owners can force DLSS 4 in unsupported games

A new Nvidia app update will allow RTX 50 GPU owners to force games to use the latest DLSS model using multi-frame generation, even if the game is not officially supported. This will be available following the release of a new GeForce Game Ready driver, which will be released alongside the launch of the RTX 5090.

Owners of previous generation Nvidia RTX 40 GPUs can also use the override feature, although not the multi-frame generation of DLSS 4. Nvidia confirmed that upcoming games will be liked Doom: The Dark Ages And Explained will start with DLSS 4 Multi Frame Generation at launch.

At CES, AMD also announced its new Ryzen Z2 processors for the Lenovo Legion Go, while Acer launched the Nitro Blaze 8 and 11 handhelds. Atari is also introducing a new handheld console called Gamestation Go at the convention.