A depiction of the difference between 14nm Vega 64 and 7nm Radeon VII chips. Note room in 7nm version for 16GB HBM2 that is included (RTX2080Ti has 11GB memory). For what its worth, the AMD HBM2 setup is said to transfer data twice as fast as RTX2080Ti DDR6. That helps compute, but is "overkill" for current games.

From what I read, the compute aspects of GPUs will become more used in games as time passes. For example, Nvidia uses Ray Tracing. AMD has indicated they will eventually use Ray Tracing (when well supported by games). However, in the shorter term, there is are methods to give similar results using (what I call) a compute function based on a new feature of Microsoft gaming drivers. Compute functions can make use of extra and extra fast memory. Just me rambling as these things are only hinted at in my reading.