NVIDIA GeForce RTX 50 Series Graphics Cards: Blackwell Architecture, Specs, Pricing, and Features
NVIDIA has unveiled its latest generation of GeForce RTX graphics cards, powered by the cutting-edge Blackwell architecture. The GeForce RTX 50 Series debuts at the end of January 2025, featuring four new cards and a host of innovations aimed at gamers, developers, and creatives alike.
Beyond the hardware, NVIDIA CEO Jensen Huang also announced updates to the GeForce NOW service, compatibility with Steam Deck, and enhancements to the NVIDIA Broadcast app during the CES presentation.
For a broader look at NVIDIA’s announcements, check out our CES coverage. If you're here for a breakdown of the RTX 50 Series GPUs, keep reading!
NVIDIA Blackwell GeForce RTX 50 Series Overview
The GeForce RTX 50 Series introduces a new standard in gaming and AI-driven performance, leveraging NVIDIA’s Blackwell architecture. Announced at CES 2025, the lineup includes:
- RTX 5090
- RTX 5080
- RTX 5070 Ti
- RTX 5070
These GPUs deliver exceptional improvements in processing power, AI capabilities, and ray tracing performance, setting a new benchmark for gaming and creative applications.
pppppppppppppppppppppppppp
Key Specifications
The Blackwell architecture incorporates:
- 9th-Generation NVENC Encoders: Enhanced video encoding performance.
- 5th-Generation Tensor Cores: Superior AI and machine learning performance.
- 4th-Generation Ray Tracing Cores: Improved real-time ray tracing for lifelike graphics.
Highlights of the Lineup
-
RTX 5090
- AI TOPS: 3,352
- CUDA Cores: 21,760
- Memory: 32 GB GDDR7
- Bandwidth: 1,792 GB/s
- Total Graphics Power (TGP): 575W
- Claimed to deliver twice the AI-enhanced gaming performance of the RTX 4090.
-
RTX 5070 (entry-level card)
- AI TOPS: 988
- CUDA Cores: 6,144
- Memory: 12 GB GDDR7
- Bandwidth: 672 GB/s
- Over twice the AI TOPS of its predecessor, the RTX 4070.
Model |
AI TOPS | CUDA Cores | NVENC | NVDEC | Memory | Bandwidth | TGP |
|---|---|---|---|---|---|---|---|
| RTX 5090 | 3,352 | 21,760 | 3x 9th Gen | 2x 6th Gen | 32 GB GDDR7 | 1,792 GB/s | 575W |
| RTX 5080 | 1,801 | 10,752 | 2x 9th Gen | 2x 6th Gen | 16 GB GDDR7 | 960 GB/s | 360W |
| RTX 5070 Ti | 1,406 | 8,960 | 2x 9th Gen | 1x 6th Gen | 16 GB GDDR7 | 896 GB/s | 300W |
| RTX 5070 | 988 | 6,144 | 1x 9th Gen | 1x 6th Gen | 12 GB GDDR7 | 672 GB/s | 250W |
Jensen Huang described Blackwell as "the most significant computer graphics innovation since programmable shading 25 years ago."
Pricing and Availability
- RTX 5070: $549, releasing February 2025
- RTX 5070 Ti: $749, releasing February 2025
- RTX 5080: $999, releasing January 30, 2025
- RTX 5090: $1,999, releasing January 30, 2025
atOptions = { 'key' : '26d0865120c9665c7b7aaf310021afdd', 'format' : 'iframe', 'height' : 90, 'width' : 728, 'params' : {} };
DLSS 4: Enhanced Upscaling and Performance
The RTX 50 Series debuts DLSS 4, a revolutionary update to NVIDIA's AI-powered upscaling technology. Powered by a transformer-based model, DLSS 4 introduces:
- Multi Frame Generation: Generates up to three additional frames per traditionally rendered frame, enhancing fluidity.
- Improved Efficiency: 40% faster performance with 30% less VRAM usage compared to DLSS 3.
At launch, 75 games will support DLSS 4, with more titles in development. The technology also introduces Reflex 2, reducing latency by up to 75% using Frame Warp technology.
Backward Compatibility
RTX 50 GPU owners can force DLSS 4 into unsupported games using Multi Frame Generation. While RTX 40 users can use some DLSS 4 features, Multi Frame Generation remains exclusive to the RTX 50 Series.
Stay tuned for more updates as NVIDIA’s RTX 50 Series redefines gaming and AI-driven applications!