OK. Tonight, decided to put back the 8600K with the 5700XT and try ESO again for sanity check.
Since I still have the old NVME drive on the Z370 board, just plugged cooler and power and run ESO at the same settings currently have it on the 3900X.
So 8600K @ 5Ghz with the 5700XT, a single core heavy game, the same town I was in had 32-48fps. On the 3900X boosting to 4.35, around 76-80fps with few dips to 55fps.
Run the same dungeon, the 8600K got 110-120 fps (the drops at multi target fights), on the 3900X stable 144fps with few drops to 141fps on 2-3 places with plenty of effects. (I have 144fps cap on drivers and the game config)
On the 3900X the 5700XT clocks were between 1980-2070 (mostly sitting at 2020), and on the 8600K 1730-1800. That's a huge discrepancy and makes no sense given that even if we add the ~10% IPC boost on clock for clock the Zen 2 got, is shy of the 5Ghz the CFL was running.
The 3900X system has a brand new MP600 drive, but the EVO 970 on the Z370 board is faster, so any benefits from loading faster the graphics doesn't count.
Tomorrow I will run the GPU to PCIe 3.0 on the 3900X, just in case is the cause of the extra fps, as the game is DX11 and on some games the benchmarks have proven that the 5700XT benefits greatly from 4.0 over 3.0.
Has anyone else tested their 5700XT/5700 switching between 3.0 & 4.0 on their DX11 games?