Today, NVIDIA introduced the GeForce RTXTM 50 Series desktop and laptop GPUs, the most cutting-edge consumer GPUs for creators, developers, and gamers. The GeForce RTX 50 Series, which is powered by the NVIDIA Blackwell architecture, fifth-generation Tensor Cores, and fourth-generation RT Cores, offers advances in AI-driven rendering, including digital human technologies, geometry, lighting, and neural shaders.
“Blackwell, the engine of AI, has arrived for PC gamers, developers and creatives,” said Jensen Huang, founder and CEO of NVIDIA. “Fusing AI-driven neural rendering and ray tracing, Blackwell is the most significant computer graphics innovation since we introduced programmable shading 25 years ago.”
With 92 billion transistors, the GeForce RTX 5090 GPU is the fastest GeForce RTX GPU to date and can process more than 3,352 trillion AI operations per second (TOPS).
The GeForce RTX 5090 GPU performs up to two times better than the GeForce RTX 4090 GPU thanks to Blackwell architectural advancements and DLSS 4.
GeForce Blackwell comes to laptops with all the functionality of desktop machines, significantly improving portable computing with its exceptional efficiency and graphics capabilities.
The Blackwell generation of NVIDIA Max-Q technology comprises tiny and light laptops that keep their sleek design without compromising performance or power, and it can increase battery life by up to 40%.
NVIDIA DLSS 4 Increases Performance by Up to 8x
DLSS 4 introduces Multi Frame Generation, which uses AI to produce up to three frames each to increase frame rates.
Together with the DLSS portfolio of technologies, it can boost speed by up to eight times compared to traditional rendering while still preserving NVIDIA Reflex technology responsiveness.
Additionally, DLSS 4 presents the transformer model architecture’s first real-time implementation in the graphics sector.
To give better stability, less ghosting, more details, and improved anti-aliasing in gaming scenes, Transformer-based DLSS Ray Reconstruction and Super Resolution models employ twice as many parameters and four times as much computation.
On launch day, DLSS 4 will work with more than 75 games and apps on GeForce RTX 50 Series GPUs.
NVIDIA Reflex 2 presents Frame Warp, a novel method for lowering game latency that updates a generated frame in real-time based on the most recent mouse input right before it is sent to the screen.
Up to 75% less latency is possible with Reflex 2. In addition to making single-player games more responsive, this gives players a competitive advantage in multiplayer games.
Blackwell Introduces Shaders with AI
NVIDIA’s introduction of GeForce 3 and programmable shaders 25 years ago laid the groundwork for two decades of graphics innovation, ranging from real-time ray tracing to compute shading and pixel shading.
NVIDIA is launching RTX Neural Shaders in conjunction with GeForce RTX 50 Series GPUs. These shaders leverage small AI networks to enable film-quality lighting, materials, and other features in real-time games.
Since people are prone to noticing even the smallest mistakes or artefacts in digital humans, rendering gaming characters is one of the most difficult challenges in real-time graphics.
RTX Neural Faces uses generative AI to produce a high-quality, temporally stable digital face in real-time using basic rasterised face and 3D position data.
New RTX technologies for ray-traced skin and hair accompany RTX Neural Faces. These developments are expected to significantly increase the realism of game characters and surroundings, especially when combined with the new RTX Mega Geometry, which allows for up to 100 more ray-traced triangles in a scene.
With Zorah, an innovative new technology demonstration from NVIDIA, the potential of neural rendering, DLSS 4, and the new DLSS transformer model is demonstrated on GeForce RTX 50 Series GPUs.
CEO Huang revealed the complete portfolio, including the RTX 5090 and RTX 5070, at CES 2025. Let’s start by discussing prices:
- RTX 5090: $1,999
- RTX 5080: $999
- RTX 5070 Ti: $749
- RTX 5070: $549.
The powerful beast that is the RTX 5090 will be the leader of the Blackwell family. Nothing unexpected there. Nvidia claims that the Blackwell series, which has 92 billion transistors, can achieve up to 125 shader TFLOPS, 380 RT TFLOPS, and 4,000 AI TOPS. However, the RTX 5090 is reported to have 3,400 AI TOPS, which is slightly less than the maximum and suggests it isn’t fully utilising the processor.
Nvidia GeForce RTX 5000: Anticipated Features and Details
With significant performance gains over the previous generation, Nvidia’s most recent Blackwell architecture is anticipated to be included in the RTX 5000 series. The RTX 5090, the flagship model, is said to have a 600-watt power standard and 32GB of GDDR7 memory.
Furthermore, a 256-bit memory bus with 16GB of GDDR7 VRAM is anticipated for the RTX 5080, whilst a 192-bit memory bus and 12GB of GDDR7 VRAM may be included for the RTX 5070.
It is expected that all models will use a single 12VHPWR connector. By improving real-time rendering and AI-driven graphics advancements, the upcoming series is also anticipated to deliver “Neural Rendering Capabilities,” which might completely transform graphics processing.
Professionals and gamers alike should expect to see notable performance improvements with the GeForce RTX 5000 series. While experts in domains like 3D rendering, video editing, and AI development might profit from more computational power and efficiency, gamers should expect faster frame rates and better visual fidelity.
The new DLSS Transformer model, which shares the same basic architecture design as AI models like ChatGPT or Google Gemini, will power Ray Reconstruction, Super Resolution, and DLAA. These can enhance motion detail, decrease ghosting, and increase temporal stability, according to Nvidia.
Regarding when you can obtain these cards, the availability is as follows:
- RTX 5090 (30 January)
- RTX 5080 (30 January)
- RTX 5070 Ti ( February)
- RTX 5070 in (February)
