Connoisseurs of Nvidia history will not be surprised by video cards whose cost is measured in several thousand dollars, but if previously these were specific products for the wealthiest connoisseurs of gaming graphics, the January update to the range of gaming solutions brought the GeForce RTX 5090 as a new mass flagship, and the company rated this video card at $1999.
Nvidia founder and permanent CEO Jensen Huang did not postpone the presentation of new gaming solutions with Blackwell architecture until the final part of his speech in Las Vegas. Shortly after he appeared on stage in yet another new leather jacket, he talked about how artificial intelligence could create more realistic 3D graphics. Then a GeForce RTX 5090 video card and a printed circuit board from it appeared in his hands, containing not only a graphics processor with 92 billion transistors, but also 32 GB of GDDR7 memory.
As promised by the head of Nvidia, the GeForce RTX 50 family will begin to conquer the market in January of this year. The junior solution, called GeForce RTX 5070, with a recommended retail price of $549, will offer a performance level that is not inferior to the GeForce RTX 4090; such a video card will go on sale in February. The next product in the family in order of increasing cost will be the GeForce RTX 5070 Ti for $749, and a step away from the flagship will be the GeForce RTX 5080 for $999. It is the GeForce RTX 5080 that will go on sale on January 30 of this year along with the GeForce RTX 5090. The GeForce RTX 5070 Ti will be offered only in the version of Nvidia partners, without a reference design, but the younger GeForce RTX 5070 will be offered in the Founders Edition.
According to VideoCardz, all four video cards of the GeForce RTX 50 family will use a 16-pin additional power connector and support PCI Express 5.0, DisplayPort 2.1b and HDMI 2.1a interfaces. This will allow you to connect monitors with a resolution of up to 8K and a refresh rate of 165 Hz. Support for FP4-precision computing allows new graphics cards to reduce the size of language models and be limited to less memory, with virtually no sacrifice in computing quality.
The flagship GeForce RTX 5090 also has three encoders and two decoders, allowing it to export video 60% faster than the GeForce RTX 4090 and four times faster than the GeForce RTX 3090 in this area. In addition, the NVENC video encoding unit itself has been improved and received higher performance. The sixth generation decoder doubled the speed of working with H.264. The fourth generation of RT cores allows the GeForce RTX 50 family to demonstrate 40% higher performance in 3D graphics for video cards. Support for DLSS 4 will provide the GeForce RTX 5090 with up to two times superior graphics performance compared to the GeForce RTX 4090.
The TGP level of the GeForce RTX 5090 video card has increased to 575 W, which is 125 W more than the GeForce RTX 4090. The recommended power supply power is 1000 W. At the same time, the overall dimensions of the flagship video card of the new family turned out to be quite modest; it will easily fit into the space of two expansion slots. In addition to 32 GB of GDDR7 memory, the new product can offer 21,760 CUDA cores and memory bandwidth of 1,792 GB/s.
The GeForce RTX 5080 will be twice as fast as the GeForce RTX 4080, while the new product will be equipped with 16 GB of GDDR7 memory with a bandwidth of 960 GB/s and 10,752 CUDA cores. At a TGP level of up to 360 W, it will require a power supply of at least 850 W. The GeForce RTX 5070 Ti will retain 16 GB of GDDR7 memory, but its bandwidth will be limited to 896 GB / s, and the number of CUDA cores will not exceed 8960 pieces. At a TGP level of no more than 300 W, the recommended power supply power will be 750 W. Finally, the GeForce RTX 5070, with 6144 CUDA cores, will be equipped with 12 GB of GDDR7 memory with a bandwidth of 672 GB/s, and a TGP level of 250 W will be combined with a power supply with a recommended power of at least 650 W.
In March, laptops based on the new generation GeForce RTX mobile graphics solutions should hit the market. Laptop models based on the GeForce RTX 5070 with 8 GB of GDDR7 memory will be offered at prices starting from $1,299; laptops with GeForce RTX 5070 Ti graphics and 12 GB of GDDR7 memory will be three hundred dollars more expensive, but for the purchase of a gaming laptop with a discrete GeForce graphics solution RTX 5080 and 16 GB GDDR7 will already require an amount of at least $2199. Finally, laptops powered by the flagship GeForce RTX 5090 will start at $2,899 and will offer 24GB of GDDR7 memory.