I'm not sure I understand the question.
You would have compressed data that you transfer then process/decompress in memory.
First an explanation. What I'm describing is what the YouTube channel XboxBr explained as beeing the way the Velocity Engine works.
SFS and its ID buffers connected to the GPU and the 100 GB virtual RAM Partition that would hold uncompressed textures as if they were in RAM.
Was the explanation correct? Dont know, and that is why I'm questioning, because if it is, I dont see it working on compressed texture data.
That is why I placed the question.
The explanation seemed logic to me since SFS is a GPU feature, and as such works with GPU memory, and this partitition is meant to to work as virtual memory, keeping data on the same state as if it was on real RAM.
As such, this can give effective gains by only supplying (reading) the required part of the the textures to the GPU real RAM.
But if the system works like that, it seems to pose a question. Can these gains be applyed to compressed data?
Because to compress data you would have to read the uncompressed data to RAM, and this would kill the need for compression since data would already be in RAM and the SSD transfer capacity already used. Besides these compressors in hardware are able to decompress in real time. But can they compress with the same efficiency? And can they compress and decompress in sequence keeping things in real time?
As I see it, the gains can only be applyed to uncompressed data. And this was why I questioned. So that if I what was explained is wrong, or if what I understood is wrong someone could explain me the process