CES 2025: NVIDIA® RTX™ 50 series presented

On January 7, at CES 2025, NVIDIA® CEO Jensen Huang unveiled the highly anticipated RTX™ series graphics cards based on the Blackwell architecture. These cards represent a remarkable digital transformation, combining traditional computing with AI in a complementary relationship. This powerful combination delivers exceptional performance not just in games and applications, but also enables local processing of complex AI models, bringing advanced artificial intelligence capabilities to everyday users.

In total, the line includes 4 models:

1. RTX™ 5070
2. RTX™ 5070 Ti
3. RTX™ 5080
4. RTX™ 5090

All cards in the series support PCIe v5.0 and can output 4K resolution at up to 480 Hz across 4 monitors (3 DisplayPort + 1 HDMI). For 8K resolution, the maximum refresh rate is 120 Hz. Let’s explore the basic technical specifications:

All GPUs feature the new GDDR7 memory standard, replacing the previous GDDR6X. The memory bandwidth varies according to the number of memory chips installed:

The NVIDIA® Blackwell GPU packs an impressive 92 billion transistors on a single silicon crystal, necessitating robust cooling. The presentation showcased a design featuring two fans and a large evaporation chamber, with heat pipes removing heat in both directions.

Presented GPUs include support for DLSS 4 technology, which works 8 times faster than traditional rendering. Game developers have long wrestled with balancing image quality and performance. Higher quality typically meant lower frame rates, while maximizing frame rates required quality sacrifices. DLSS offered a novel solution by using AI algorithms to generate additional frames, boosting frame rates without heavily loading the GPU or compromising quality.

As many experts predicted, support for 5th-generation Tensor cores with new number formats has appeared: FP6 and FP4. This significantly speeds up the work of large neural network models and generative AI. Ray Tracing cores have been upgraded to their 4th generation. Working in tandem with Tensor cores, they enable developers to revolutionize traditional rendering approaches through AI-powered texture compression. This results in up to a sevenfold reduction in VRAM and system memory usage.

For gamers, the RTX™ 50 series introduces Frame Warp: a technology that builds upon NVIDIA® Reflex 2. Frame Warp reduces game latency by updating rendered frames with the latest mouse input just before display. This technology cuts latency by up to 75%, enhancing both multiplayer competition and single-player responsiveness.

The RTX™ 50 series will likely become the fastest and most promising GPUs in the consumer market. With its advanced architecture and refined manufacturing process, this generation promises to deliver exceptional performance gains across gaming, content creation, and AI workloads, making it a potential game-changer for enthusiasts and professionals alike. Here, on LeaderGPU, we’re getting ready to test these impressive GPUs and look forward to sharing our findings.