as I have stated in the title I would like to know if this new methodology of compressing the textures can be used in videogames to provide a very good visual quality while consuming a low amount of vram (generally I’d say 8 or 12, also if this is possible how should this feature be scaled on different hardware like high end rtx 4090-4080 and low end hardware such as the rtx 3060 ti?
Hello @micheleiadarola03 and welcome to the NVIDIA developer forums.
One of the premises in the introduction for the SIGGRAPH publication on neural texture compression is the ever increasing need for high definition textures. As such it would be only natural to apply this to reduce the amount of GPU memory necessary to store textures.
If this will be done in practice in the future and how that would be scaled to different Board designs is very much up to the game developers making use of this new approach. It could just as well be that they decide to put more textures into the saved memory instead of reducing the memory footprint.
I also recommend to have a look at the other SIGGRAPH papers presented in one of our last Blog posts.
Thanks!
I guess that time will tell.
I hope that during the siggraph you can show something more about this new technology because for Unreal Engine 5 we need a lot of memory and 8gb are not enough.
1 Like