Jensen Huang walked onto the CES 2025 stage in Las Vegas and held up a blacked-out graphics card. “Here it is, our brand-new GeForce RTX 50 series, Blackwell architecture,” he told an audience of more than 6,000. “The GPU is just a beast.”
The card was the RTX 5090, which Nvidia calls the most powerful GeForce GPU it has made. It packs 92 billion transistors and is rated at 3,352 trillion AI operations per second. Huang pointed out that it gets by with two cooling fans. “Even the mechanical design is a miracle,” he said.
Release dates
The RTX 5090 and RTX 5080 desktop cards go on sale on January 30. The RTX 5070 Ti and RTX 5070 follow in February, and laptop versions are due in March.
DLSS 4 and the software side
Much of the performance pitch rests on software. DLSS 4 adds Multi Frame Generation, which Nvidia says can lift performance by up to 8x. Huang described it as producing three extra frames for every one the card calculates the traditional way. “We’re able to render at incredibly high performance, because AI does a lot less computation,” he said.
Nvidia also announced Reflex 2, which it says can cut PC latency by up to 75%.
Other graphics features got a mention. RTX Neural Shaders use small neural networks to improve textures, materials and lighting during play. RTX Neural Faces and RTX Hair handle real-time faces and hair. RTX Mega Geometry raises the number of ray-traced triangles by up to 100x.
AI on the desktop
The new cards support FP4 computing, which Nvidia says can double AI inference speed while using less memory than earlier hardware. Nvidia is also offering AI foundation models for RTX PCs, packaged as NIM microservices, along with blueprints for building digital humans, podcasts, images and video. “AI PCs are coming to a home near you,” Huang said.
“GeForce enabled AI to reach the masses, and now AI is coming home to GeForce,” he said.
The 90-minute keynote went well beyond graphics cards. Huang also introduced the Cosmos platform for robots and autonomous vehicles, a partnership with Toyota on DRIVE AGX, and Project DIGITS, a compact desktop AI supercomputer built on the GB10 Grace Blackwell Superchip and set to launch in May.
