It's no coincidence that such a number of CUDA cores appeared in the RTX 4090 model presented as an April Fool's joke. Moving to a lower technological dimension should therefore translate into a significant increase in performance and better energy efficiency. The full AD102 chip could have as many as 18,432 CUDA cores (almost twice as many as in RTX 3090 and RTX 3080's GA102), and its single-precision computing performance would be expected to reach a monstrous 66.4 TFLOPS at 1800 MHz (impressive compared to GA102's 38.7 TFLOPS). She is considered the first female programmer - she became famous for developing a computer algorithm for Charles Babbage's analytical machine.

Hopper is named after Ada Lovelace, a prominent British mathematician and daughter of poet Lord George Byron.

Here's Nvidia's official roadmap until 2025 - we're supposed to get two more generations of GeForce RTX graphics cards (presumably the 40) within three years.Īccording to earlier, although unofficial reports, the successor to the Ampere architecture - used in Nvidia's current generation of GPU - is to be produced in a 5-nanometer technological process Nvidia Lovelace (instead of Nvidia Hopper, which in turn will go - according to the latest roadmap - to SoC ARM units designed for advanced calculations associated with artificial intelligence, among other things).

Nvidia geforce mx 4000 driver update