by Jordi Bercial Yesterday at 17: 41 1

As we can read in WCCFTech , it seems that the assumption delay of Hopper GPUs with MCM architecture is not going to be a problem for the advancement of NVIDIA in order to create large GPUs, and it is that as we see , the company is going from strength to strength to create an even bigger and more powerful GPU than Ampere.

In figures, we would find ourselves before a GPU with 12 GPCs and 72 TPCs that would host 144 Stream Multiprocessors for a total of 18432 CUDA nuclei , figures that exceed the already enormous figures of which the GPU Ampere GA 102, with 7 GPCs, 41 TPCs and 84 Stream Multiprocessors for a total of 10752 CUDA cores, respectively.

Kopite7kimi confirmed this data extrapolated by 3DCenter, so It is to be expected that, except for changes or inconsistencies, these data will be the ones that reach the top-of-the-range GPU of the architecture NVIDIA ADA Lovelace, which would belong to a possible NVIDIA RTX 4090 or its server counterpart.

Be that as it may, although it is not the NVIDIA Hopper architecture that some rumors promised us regarding the launch of a GPU with chiplet design , it does seem to be an extremely powerful chip, so we will have to be attentive to future leaks.

End of Article. Tell us something in the Comments or come to our Forum!

Jordi Bercial

Avid enthusiast of technology and electronics. I messed around with computer components almost since I learned to ride. I started working at Geeknetic after winning a contest on their forum for writing hardware articles. Drift, mechanics and photography lover. Don’t be shy and leave a comment on my articles if you have any questions.