NVIDIA RTX 5090 Review Before Release in December 2025

NVIDIA RTX 5090 Debut and Technical Leap

At the beginning of 2025, the graphics technology market will undoubtedly be shaken by the new flagship of the „GeForce” series — NVIDIA RTX 5090. The company has confirmed that the official RTX 50 series presentation is scheduled for the CES 2025 show in January, where alongside the most powerful model, the RTX 5080 and 5070 versions are also expected to be unveiled. Although official information is still limited, reliable sources and leaks paint an extremely ambitious picture: the card will be based on the „Blackwell” architecture, integrating fourth-generation ray tracing and fifth-generation tensor cores.
The technical specifications promise uncompromising power: it is expected to feature as many as 21,760 CUDA cores and impressive 32 GB GDDR7 memory. This is not only double the VRAM of its predecessor (the RTX 4090 had 24 GB), but also a qualitative leap in data transfer — a 512-bit bus will allow for approximately 1.79 TB/s bandwidth. However, this power will come at the cost of energy consumption: the TDP is projected to be around 575 W (about 125 W more than the RTX 4090), meaning users will need to ensure they have extremely powerful power supplies.

Does the Price-to-Power Ratio Meet Expectations?

Technological Gap Between RTX 4090 and 5090

Compared to the current market king, the RTX 4090, the new model offers more than just an evolutionary upgrade. The RTX 5090 is expected to feature 33% more cores, a third more video memory, and nearly 1.8 times greater memory bandwidth. This is further enhanced by the transition to the PCIe 5.0 standard and an updated NVENC encoder for even more efficient video processing and streaming.

The „Giant” Price and Availability

However, such engineering advancements come at a cost. Market analysts estimate that the starting price of the RTX 5090 could be around $2000, which is a significant increase compared to the starting price of the 4090 (~$1599). Experts have already dubbed this model a „giant” — it is a product for enthusiasts who require 32 GB of memory and immense computational power, but for many average users, such specifications (and power consumption) may be excessive.

FPS Metrics, DLSS 4, and the Impact of the „Blackwell” Architecture

Raw Performance Without Artificial Enhancements

Initial leaked performance tests show that even without software enhancements, the RTX 5090 demonstrates 30–40% higher performance at 4K resolution compared to its predecessor. In the game „Black Myth: Wukong,” a ~38% increase in frame rate (FPS) is recorded, while „Cyberpunk 2077” runs about 35% faster at maximum settings. More conservative estimates suggest that the typical increase in „raw” rasterization power will be around 20%, but that is only part of the story.

DLSS 4 and Frame Generation Revolution

The real breakthrough is associated with DLSS 4 technology and its frame generation capabilities. In some scenarios, this AI system can double the frame rate, allowing the new architecture to achieve results that are physically impossible for older generation devices. This confirms the trend that future graphics will increasingly rely on AI-generated image enhancement, rather than just the capabilities of the hardware.

Why This is More Than Just a Gaming PC Component

The RTX 5090 aims to become an indispensable tool for professionals working in video editing, 3D graphics, and artificial intelligence.

• Video Editing: In programs like „DaVinci Resolve,” a 10–20% overall performance increase is expected, while in specific GPU-intensive effects, it could be as high as 50%.
• 3D Rendering: In tests with „V-Ray” and „Blender,” the new card outperforms its predecessor by 35–38%, which directly translates into saved working time on large projects.
• Artificial Intelligence (AI): Here, the biggest leap is recorded. With 680 tensor cores (up from 512) and new precision standards, processing large language models (LLMs) is accelerated by about 72%, while computer vision tasks see a speedup of about 44%.

AMD Strategy and the Direction of the „Radeon RX 8000” Series

It seems that in the near future, the NVIDIA RTX 5090 will have no direct competitor. AMD, with its „Radeon RX 8000” (RDNA 4 architecture) series, has chosen a different path — „Team Red” representatives publicly declare their intention to target the mid-range market, rather than compete for the premium crown. Rumors about the upcoming RX 8700 XT (Navi48 chip) suggest that AMD will offer around 56 compute units and 16 GB of memory. This significantly lags behind the RTX 5090’s specifications and even the current RX 7900 XTX. While there is talk of a potentially more powerful AMD model with 32 GB of memory, it is likely to be aimed at data centers, leaving NVIDIA as the undisputed leader in the high-end gaming segment.

Numbers That Define the 2025 Standard