Adaptation of AMD's Fluid Motion Frames for Steam Deck

[pogofe]

With the source code of FSR3 out, it shouldn't be hard to implement it at system level just like FSR1 works and how HYPR-RX works with AMD.

The idea here is to implement Frame Generation to a specific performance target just to reach the minimal gameplay experience, It would work a bit different from the usual FSR 3.

For example, if the game is running at 23 frames per second, and the target is 30 frames per second, the algorithm would only generate 7 frames instead of trying to double the framerate to 46 frames per second.

So this way, the performance target would be reached and the input lag wouldn't be heavily affected like it happens when you use FSR3 with lower base framerates, fitting well the case of use for the steam deck.

That would keep the steam deck able to run heavy games.

[TheGamingHD]

I really like this idea but as a game programmer myself, this is very unlikely to happen in the near future. FSR 2 and 3 require information that can only be obtained in games such as motion vectors & separate rendering of game and HUD. This means it has to be integrated on a per game basis.

Despite this, It would be a really neat idea if they could do this or something similar.

Maybe in the future game engines would automatically give this information to the graphics drivers / SteamOS and maybe we could see something like this happen.

What I'm trying to say is that we need a standard way to communicate this between the OS & the game.

We would basically need this to happen between games and SteamOS:
SteamOS: "Hey game, what FSR modes do you support?"
Game: "I support FSR 1, 2 and 3"
SteamOS: "Cool can we enable FSR 3?"
Game: "Sure thing"

[TheGamingHD]

Also while AFMF is driver based and would be more feasible and easier to implement, it does not currently have a Linux implementation that Valve could implement. It is also an inferior version of FSR 3 overall due to the lack of said information from the game engine. (This is why AFMF turns off with movement on screen)

[Dakar_131]

Seria genial la incorporación de esta tecnologia como extra en el apartado de personalización de juego, no veo por que no lo hacen, se veria genial, dado el ejemplo que da "lossless scaling" pueden hacerlo de igual manera

[Nobbie]

I wouldn't be surprised if it gets linux support soon seeing how big linux is growing lately

[Prezidentas]

Originally posted by pogofe:

For example, if the game is running at 23 frames per second, and the target is 30 frames per second, the algorithm would only generate 7 frames instead of trying to double the framerate to 46 frames per second.

It doesn't work like that. Inserting a fake frame in between two real frames is easy. How are you going to time 7 frames into 23 FPS game? The problem is frame timing - the only way to do it smoothly is to double the framerate.