AMD and NVIDIA dispute a war away from hardware which we are all used to. Obviously, this constant push and pull to gain more market share in graphics cards has been making headlines for years, but they both realized that it’s not just the numbers, it’s the users you manage to retain.
That is why both set out to develop unique technologies and, in the case of the scaling in video gamesNVIDIA stepped on the accelerator with its DLSS. AMD does not want to stop fighting and although its first attempt to kill NVIDIA’s “Deep Learning Supersampling Technology” did not go as they wanted, AMD FSR 2.0 comes to try equate things.
The update was a half-truth, but it was not until today that we had confirmation from AMD of the second generation of its FidelityFX Super Resolution technology. In this case, the company has redone the code from scratchalthough it continues without using machine learning via artificial intelligence, as NVIDIA does. To match the technology to DLSS, FSR 2.0 will use a temporary scaling instead of spatial scaling as its first version did.
The difference between the two is that while spatial scaling improves the image based on the incoming data, that is, the information on the screen. In the case of temporal scaling, it uses data from frames above to further improve the current image quality.
On the other hand, AMD wants to distance itself from NVIDIA’s neural network that has given such good results with DLSS 2.0 and advances in its statement that this technology It can be used on both AMD GPUslike NVIDIA, without neglecting the most Recent Intel ARC. The “red team” dispenses with an exclusivity similar to DLSS, although they say that it is not clear if some hardware improvements will be limited to products signed by AMD in the future.
As for the comparisons and gross performance, the Californian company has shared a video with the improvements in Deathloop, the Arkane title, but it has an ace up its sleeve pending the imminent GDC on March 23. However, we have already been able to see some promotional images comparing it only with its previous version and a launch window: second quarter of 2022.