There is not much time left until Nvidia finally reveals the next generation of GPU. In the meantime, most of the information leaked on the internet has helped us get a little idea about it; but nothing concrete about performance, the raw power that hides behind the green curtain. The company is working hard on its new GPU Ampere A100, which is intended to be the fastest creation of the company so far. The graphics card is manufactured on 7nm incorporating 54 billion transistors, 6912 CUDA cores, and 40GB of super-fast memory; of the HBM2E type with a bandwidth that can reach 1555 GB/s.
How fast do we expect it to be more accurate? Well, thanks to Jules Urbach, CEO of OTOY, the developer and creator of the OctaneRender software, we got the first benchmark of the Ampere A100.
The beast managed to reach a score of 446 points in OctaneBench, the benchmark made especially for OctaneRender; surpassing the fastest GPU today. As a benchmark, the GeForce RTX 2080Ti scores 302 points, with the Nvidia A100 Ampere being up to 47.7% faster than the Turing.
However, the fastest Turing board found in the benchmark database is the Quadro RTX 8000; the GPU dedicated to professionals who obtained 328 points, showing that the Turing architecture is still very well maintained.
Most importantly, the Nvidia Ampere A100 ran the test with RTX technology disabled, depending on the potential of the Ampere architecture and the quality of the silicon, the end result may vary. I am very curious about the next generation of graphics cards, is it possible that this generation will be a total fail? Or something completely new? Remains to be seen.
