NVIDIA 7nm Ampere GPU in the first benchmarks, the results are exceptional

As can be seen from the screenshots, the software does not recognize the model of these GPUs, however detecting Streaming MultiProcessor, amount of memory and GPU frequency; the data obviously must be taken with due caution, however, looking above all at the memory cut, it seems clear that we are facing two high-performance GPUs aimed at the professional segment.

Without the necessary premises, the more powerful of the two NVIDIA GPUs (the 118 SM one) marks an impressive 184,096 in the GeekBenck test, about 40% more than a TITAN RTX. This chip should therefore offer 7552 Cuda Core (64 cores for SM) and 24GB of memory (GDDR6?) With a GPU frequency of 1.1 GHz. The second model, equipped with 108 SM, will therefore have 6912 Cuda Core, while it is not clear the amount of memory, detected by Geekbench in even 47GB. In this second case, the GPU clock drops to 1.01 GHz with a GeekBench score of 141654, slightly higher than a TITAN RTX.

NVIDIA at the moment has not commented on the news, we are awaiting further updates to be able to confirm first of all the technical specifications of these two brand new chips that, from our point of view, to support such a large number of SM are likely to use the new production node at 7nm.