An imposing design
The size of the RTX 5090 cards was already the subject of much speculation before CES, and the leak of an image of the Inno3D iChill X3, relayed by the Videocardz site, only confirmed these fears. The RTX 5090 is 3.5 slots deep, a size equivalent to the larger RTX 4090, but its chassis appears even wider. This risks complicating its integration into the most modest cases, and in particular for mini-ITX configurations.
Nvidia seems to want to maintain this massive approach with its Founders Editions and those of its partners. The impressive cooling also underlines the power of the card, but can pose a problem for users looking for a minimum balance between performance and compactness.
Specifications that set a new standard
On the technical side, the RTX 5090 displays impressive characteristics. With 32 GB of GDDR7 memory, it far exceeds the 24 GB of GDDR6X present on the RTX 4090. This 8GB increase is the most notable development in the RTX 50 lineup, particularly useful for memory-intensive tasks like AAA games and 3D creation applications.
The Blackwell graphics processor, the heart of this new generation, integrates 92 billion transistors. This also translates into a computing power of 3,352 TOPS (Tera Operations Per Second) dedicated to artificial intelligence operations. Nvidia thus promises performance gains reaching twice those of the RTX 4090, in particular thanks to the fourth generation of RT cores and the fifth generation of Tensor cores.
DLSS 4 and innovations in artificial intelligence
Among the key new features, DLSS 4 plays a central role. This technology relies on artificial intelligence models capable of generating up to three additional frames for each frame rendered by the GPU. According to Nvidia, this would allow a performance improvement of up to 8X, while maintaining visual quality close to the original.
DLSS 4 leverages real-time transform models, which provide significant improvements in detail, stability and artifact reduction. In addition to this, RTX Neural Shaders technology sets up neural networks within the shaders themselves, which opens the way to renderings close to cinematic productions.
Nvidia also offers new solutions to make digital faces more realistic with RTX Neural Faces, and advances in hair and skin rendering for even more immersive environments.
A chilling price
If the features impress, the price of the RTX 5090 will calm everyone down. Sold at $1,999 for the Founders Edition, this card is positioned well above the $1,599 of the RTX 4090 upon release. The other models in the series follow the same trend: $999 for the RTX 5080, $749 for the RTX 5070 Ti and $549 for the RTX 5070.
This price increase could limit the adoption of these new cards by the general public, even if Nvidia is betting on the interest of professionals and enthusiasts to compensate for this obstacle.
Compatibility and energy consumption
Another question mark remains the power consumption of the RTX 5090. Although no official information has been communicated on this subject, rumors suggest the use of a 16-pin 12VHPWR connector, already present on the previous generation. This choice could simplify the installation, but would require a power supply capable of handling high consumption peaks, a factor to consider for users who would like to upgrade their system.
Availability and soft launch
The Nvidia RTX 5090 will be available from January 30, 2025, with a simultaneous launch of Founders Editions and models customized by brands like ASUS, MSI, Gigabyte or Inno3D. Other models, like the RTX 5070 and its laptop variants, will arrive later in the year.
The RTX 5090 therefore marks a new stage in the evolution of GPUs, both in terms of its performance and its massive use of artificial intelligence. Its high price, its imposing dimensions and its energy requirements make it an option clearly intended for a very specific audience. It remains to be seen whether the market will follow or whether players will turn to more affordable alternatives.
