I was just thinking, with the 7900X3D/7950X3D CPUs if you run a fully multithreaded benchmark, such as Cinebench R23, on them, are the scores consistent each time or do the score vary more than 5% each time. Not all blocks take the same time to finish rendering in Cinebench as some blocks (the less complex ones) render way faster than others (more complex ones) and I wonder if each time Cinebench is run those same blocks get rendered by a different CCD than the previous run. The reason I'm wondering this is because one CCD is different than the other, and if sometimes more data is inside the 3D cached/lower clocked CCD or if sometimes more data is inside the normal/higher clocked CCD.
Another thing, If you were playing a hypothetical game that heavily utilized 12/16 cores, would the performance be more consistent if both CCD's were normal higher clocked CCD's (or both CCDs were 3D Cached lower clocked ones which such a CPU does not exist) than one CCD being the normal higher clocked one and one CCD being the 3D cached lower clocked one? In other words, in that hypothetical game, the next time it's being played is it possible that a scene that was processed in the 3D cached CCD previously is now processed in the higher clocked normal CCD the next time it is being played, maybe leading to a significant drop in performance in that specific scene if that scene heavily benefited from the 3D cache? Remeber this is assuming that this hypothetical game is heavily using 12/16 cores on the 7900X3D/7950X3D.