The current generation of video cards has split the market in two. On the one hand there is NVIDIA, which with the RTX GPUs has brought revolutionary technologies such as Ray Tracing to the public, also offering a first taste of the application of deep learning and Tensor Core with the DLSS. In their first version, however, the performance leap was not so marked compared to the previous Pascal architecture in terms of fps generated, so much so that only with the Super range has there been a greater gap between the two.
On the other hand, AMD has had to recover the lost ground in recent years, however, starting from the low and medium range and leaving the high one to NVIDIA.
Now however, with the development of the Ampere and RDNA2 GPUs we could assist to the biggest performance leap seen in recent years, thanks to the high number of calculation units inside them.
More power for everyone
All the rumors so far about NVIDIA's Ampere architecture and AMD's RDNA2 agree: the new video cards will have a large number of computing units, especially in the top models. The RTX 3060, according to the first rumors, will have 2560 Cuda Core, exactly as in the current RTX 2070 Super, but with twice the RT Core, 80 in all. Always according to the first leaked information, which we remember are still to be verified, the RTX 3080 Ti will have 8192 Cuda Core, almost double that available in the current RTX 2080 Ti, and 256 RT Core, the quadruple of the current model.
If you observe the rumors in the AMD house, more nebulous than NVIDIA, it seems that the American house is working on a high-end GPU, renamed for now "Big Navi"In this case the information is less precise but it seems that it can integrate 80 Compute Units, for a total of 5120 Stream Processor, double the current RX 5700 XT.
The AMD lineup for 2020 is more uncertain, we do not know how many models are in development and we do not know if AMD will be able to compete with the numbers leaked so far by the NVIDIA front, but also in this case we are facing potentially doubled performance compared to to the current generation.
This is without taking into account the improvements in the RDNA2 and Ampere architectures, which can lead to a higher IPC and consequently to performances of the single computing units higher than 50% in comparison to the current ones.
It is therefore clear that the element in common between the two architectures will be the high number of CUDA Core and Stream Processors available, which will be reflected throughout the lineup thanks to the high scalability of the new graphics chips. Take for example NVIDIA, which could end up with a high end composed of the RTX GPUs, designed for the 1440p and 4K with active Ray tracing, and with a medium and low composed of the GTX range, without Ray Tracing but capable of managing the best 1080p and 1440p. If the specifications leaked so far will be confirmed better forget the nomenclature used so far, the RTX 3060 will not be the heir of the current model, it will cost more and will appeal to a more demanding audience, who play at higher resolution than Full HD .
The same reasoning can be done for AMD GPUs: starting from the high number of computing units of the top model, you can create high-performance medium and low-end video cards, perhaps without Ray Tracing but with enough power to be able manage Full HD in the best way.
Ampere and RDNA2 can open a new era of PC gaming, both on the performance and price front, thanks to them great scalability in the number of computing units. Obviously we are thinking about the rumors leaked so far, far from certain, but observing the work done on the Xbox Series X GPU, which includes 3328 Stream Processors inside, the scenarios we talked about in this article become much more credible, for a 2020 that could be a great year to switch video cards.