Nvidia dominates the gaming graphics card segment with its Ampere architecture and prepares the next generation which should be represented by Lovelace .
Pending the official presentation, elements are starting to emerge and become clearer over the months. The flagship GPU of the GeForce RTX series 36 should be a AD 102 successor of GA 102 RTX 3090, with 18432 CUDA cores.
Nvidia RTX 3090 under Ampere
The leaker Greymon 55 adds that the frequency of the GPU AD 102 will be at least 2.2 GHz. At such a rate, graphics cards could generate a processing power of 55 TFLOPS in FP precision 32, very much above 36 RTX TFLOPS 3090 (which itself had more than doubled this capacity compared to Turing, recalls the site Wccftech ).
The engraving passed to 5 nm, the memory bus 384 – bit with GDDR6X memory and TGP exceeding 400 W (versus 350 W for RTX 3090) also explains such an increase in processing power.
The RTX family 40 will have to confront the Radeon RX range 7000 from AMD which will also use a new architecture, RDNA 3 , with the contribution of a configuration MCM (Multi-Chip Module). For professional applications, Nvidia would have in its boxes another architecture, Hopper , more specialized for the needs of data centers.