Specifications of Nvidia's next-generation RTX 50 series graphics cards exposed.

Nvidia is preparing to launch the next-generation Blackwell architecture RTX 50 series graphics cards. It is expected that the performance will be greatly improved. So far, there have been many rumors about GPU configurations, possible SKU lists, and expected release dates. Recently, the specifications of GeForce RTX 5090 and GeForce RTX 5080 have been further exposed.

The GeForce RTX 5090 adopts the Blackwell architecture with the code name GB202-300-A1. It is expected to be based on the GB202 graphics processor with 21,760 CUDA cores instead of the previously rumored 24,576 cores. It is equipped with 32GB GDDR7 memory with a 512-bit memory interface bandwidth and has a power consumption of up to 600W.

The GeForce RTX 5080 adopts the Blackwell architecture with the code name GB203-400-A1. It has a power consumption of 400W (the GeForce RTX 4080 series is 320W). It is equipped with 10,752 CUDA cores, 16GB GDDR7 memory, and a 256-bit memory interface bandwidth.

Overall, the performance of the RTX 5080 is almost half that of the RTX 5090. It is still uncertain why Nvidia decided to establish the next-generation product line in this way. But it is speculated that Nvidia may want to use this to limit certain consumer-level graphics cards as alternatives to AI graphics cards. However, due to the significant increase in GDDR7 memory bandwidth, the RTX 5080 should still be significantly better than its predecessor in high resolution.

If explained from a technical perspective, the GB202 processor may be composed of two GB203 processors. In other words, the RTX 5090 is very much like a dual-chip version of the 5080.