By the end of this year AMD promises to release new generation of GPU to be built on GPUs 2X Navi with the architecture of the RDNA 2. Recently, the Network appeared quite a lot of leaks regarding the characteristics of the upcoming flagship GPU, and here we all gather together.
Several fairly reputable news sources reported that the flagship GPU of AMD’s new generation, also known by the names Big 21 Navi and Navi will have 80 compute units (Compute units, CU). If the number of stream processors on the CU will not change compared to past generations, it means a 5120 stream processors.
For comparison, the older AMD GPU of the current generation – Navi 10 has 2560 stream processors. In past generations were present chips with up to 4096 stream processors. So in the next generation of AMD seems to step over this threshold, and will offer something really new.
The clock frequency of the future Navi 21 is that nothing is reported. However, if we assume that it will work on 1700 MHz, the peak theoretical performance of such a GPU will be 17.5 teraflops. Given that future GPUs will be made for 7-nm process technology, problems with the conquest of this frequency should not occur. All current 7-nm graphics chips and AMD to reach higher frequencies.
All this gives hope that soon we will see the first true flagship graphics accelerator from AMD that can compete with the most productive solutions NVIDIA. Let’s hope that it will be new decisions Ampere and not withdrawn two years ago, Turing.