nvidia now can do upsampling. Why cannot phoenix simulate general movement of low res grid and then use that data to move high res grid cubes around low res cubes. Hopefully I explained it clearly
How do you know we don’t do this already 
Seriously - this is the kind of stuff that happens internally, so there won’t be a user option for it. Phoenix does a hybrid between this method and the classical grid layout. I have some ideas how we can expand on this, but they do not match what other software have been doing so far.
Cheers!