NVIDIA said the RTX 50 series was the first consumer GPU family to support FP4, a lower-precision format. The company also said that support would let generative AI models run locally with a smaller memory footprint than on previous-generation hardware. For an independent creator, those are two separate parts of the pitch: faster image generation and less memory used by a model. [S1]

The desktop cards carried a range of announced prices. NVIDIA said the RTX 5090 would be available January 30 for $1,999 and the RTX 5080 the same day for $999. It said the RTX 5070 Ti and RTX 5070 would start in February at $749 and $549, respectively. These were announced prices and availability dates as of January 6, 2025. [S1]

NVIDIA also said it would release NIM microservices and AI Blueprints for RTX AI PCs. The company described image generation as one use case and said the microservices included the components needed to run AI on PCs. That software pipeline was presented as a future release alongside the hardware announcement. [S1]

Source documents

  1. NVIDIA Blackwell GeForce RTX 50 Series Opens New World of AI Computer Graphics | NVIDIA Newsroomhttps://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics