NVIDIA introduces GeForce RTX 5090 and 5080 GPUs with enhanced AI capabilities
From NVIDIA: 2025-02-05 09:00:44
NVIDIA introduces GeForce RTX 5090 and 5080 GPUs based on Blackwell architecture, offering 8x faster frame rates with DLSS 4, lower latency with Reflex 2, and enhanced graphical fidelity with RTX neural shaders. These GPUs support up to 3,352 AI TOPS for AI enthusiasts, gamers, creators, and developers.
NVIDIA unveils NIM and AI Blueprints at CES, designed to accelerate generative AI workloads on RTX PCs. NIM microservices provide prepackaged AI models for easy access, while AI Blueprints offer reference projects for creating cutting-edge AI experiences on GeForce RTX 50 Series GPUs.
NIM microservices simplify AI model development for PCs by offering community-driven and NVIDIA-developed AI models that are easy to download and connect to via standard APIs. These microservices are optimized for RTX GPUs and support flexible deployment options across platforms, including PCs, data centers, and the cloud.
NVIDIA collaborates with Microsoft to enable NIM microservices and AI Blueprints for RTX on Windows Subsystem for Linux (WSL2), allowing developers to efficiently build, test, and deploy AI models across platforms. NIM and AI Blueprints leverage Blackwell architecture’s innovations, including fifth-gen Tensor Cores and FP4 precision.
Blackwell architecture’s fifth-generation Tensor Cores in GeForce RTX GPUs deliver up to 3,352 AI TOPS for handling demanding AI tasks and multiple models simultaneously. This drives faster AI-driven experiences, paving the way for innovation in gaming, content creation, and beyond.
FP4 quantization format optimizes AI performance by reducing model sizes, enabling faster runtime with reduced memory requirements. FP4, native to Blackwell architecture, allows models to run faster and leaner without compromising quality, supporting efficient AI processing on local PCs for faster, smarter content creation experiences.
NVIDIA AI Blueprints, powered by NIM microservices, offer optimized reference implementations for developing advanced AI projects on RTX PCs. At CES, NVIDIA showcased PDF to Podcast blueprint, converting PDFs into interactive podcasts with AI models. Users can experiment with and develop AI projects quickly on RTX PCs and workstations. The latest AI advancements are now accessible on RTX PCs with NIM microservices and AI Blueprints. Developers and enthusiasts can experiment, build, and deploy AI locally. Initial hardware support includes GeForce RTX 50 Series, RTX 4090, RTX 4080, RTX 6000, and RTX 5000 GPUs, with more to come in the future.
Read more at NVIDIA: GeForce RTX 50 Series GPUs Power Generative AI