What NVIDIA calls DLSS 3 is actually several different technologies, one of which is called DLSS Frame Generation. It is a technology that is not new, nor is it exclusive to the green brand. Moreover, everything indicates that it will be implemented in the FSR 3 that AMD recently announced. However, what many people don’t know is that it’s possible to combine the two technologies so that all games end up running at a stable 60 FPS.
The problem of cinematic games
When choosing the graphical aspect of a game, there are two different approaches which will depend on the genre the player is playing. Specifically, if we talk about titles in genres like speed, one-on-one combat, first-person shooters. MOBA and even MMO is when the important thing is to get the highest possible frame rate. Thus, the visual quality in games is reduced due to the fact that what is important is the frame rate or rather, the game’s response to every action is immediate.
On the other side of the scale we have the games that we could consider cinematic, these base all their strength on the visual presentation, taken care of in every detail. That’s why its frame rate is lower. Unfortunately, it’s been 15 years since flat screens were standardized and we still have to support games that run at 30 FPS. Who had an explanation about TVs, not monitors, the tube.
Based on the fact that all current monitors and televisions operate at 60Hz, with 120Hz modes due to the characteristics of HDMI. It might be time to move to a higher frame rate. Some talk about 40 FPS being a multiple of 120, and there are even games under this mode. However, the technology to achieve 60 FPS in all games already exists.
NVIDIA DLSS + AMD FSR frame generation
Although NVIDIA was the first to put it on the table as one of the parts of its third generation Deep Learning Super Sampling, the part called Generation of DLSS frames It is not a deep learning algorithm and it is not an oversampling or upscaling algorithm, but it is a frame interpolation algorithm. Where from the information of two frames generated by the GPU a third is created.
The idea of generating images from other people’s information is not new and has been with us for years, in particular it is a technology used in televisions for continuous and non-interactive content. However, it is not applied in video games due to the fact that the power needed to do it is very high due to the limited time available for it. Since in a game every frame should be generated from scratch and not in a video.
Well, one of the special things about DLSS Frame Generation is that it not only gets along great with other parts of DLSS 3, but also with competing upsampling algorithms like AMD’s FSR or XeSS from Intel. That is, not only can we achieve higher frame rate through AI, but they can also be generated through interpolation.
Its application in games is already possible
Some games like the PC version of Spider-Man Remastered have shown that it is possible to combine DLSS Frame Generation and FSR from AMD with surprising results, Dwhere with an RTX 4090 and an i9-12900K it was possible to go from 125 FPS to 220 FPS on average. Obviously this speed is not necessary, but it was achieved without visual artifacts of any kind and also without problems between the two algorithms.
Which bodes well for the future for smaller systems and games that typically run at 30 or 40 frames per second. Which will be able to maintain their visual quality and reach a stable 60 FPS without any problem. Additionally, at these speeds, the power of a high-end graphics card is not needed for frame interpolation and scaling. However, it has a downside, which is the increased response latency versus frame rate, because it’s unbalanced, although that’s the point that separates the most cinematic games from the rest, they’re designed for a longer duration. answer.
Will this apply to all games or just a few?
We are very clear that FSR 3 will integrate frame interpolation and the next version of Intel XeSS too, mainly not to depend on NVIDIA’s solution. Although we have to start, although frame interpolation is accelerated by using the same type of units as Deep Learning, it is not such an algorithm. So you don’t need any kind of training in games. It is therefore something that can greatly benefit everyone, regardless of the time they were released and whether they are on PC or emulators.
The good of it all? We will be able to get rid of the scourge of games at less than 60 FPS once and for all and make them synchronize with the speed of our monitor. For the moment and while waiting for new solutions to come out, we will have to settle for the combination of DLSS Frame Generation for the interpolation, which for the moment only works on NVIDIA RTX 40. But it is a first step to solve a long-standing problem.
Regarding its application on consoles, since the RDNA 2 architecture used in PS5 and Xbox Series lacks the necessary units to accelerate the execution of these algorithms. It is very likely that only PC users have this privilege. Moreover, in the face of emulators. It should be considered a luxury to be able to play games of yesteryear at frame rates they didn’t and shouldn’t have had.
Table of Contents
