I don’t know all the technical stuff on what exactly is happening with the reset or how GPUs really works, but since this bug was first reported in 2012, the same generation that used the GCN architecture, and the card that has the most reliable patch is Navi (which uses a hybrid GCN/RDNA architecture), is there a chance that the reset bug is a GCN specific bug?
Of course it’s too early to tell since Big Navi isn’t out yet, but is that a crazy thought that it’s an architectual issue rather than card specific?
The bug is the result of improper programming, not the arch.
The bug remains present to this day because there is little incentive to fix it (BIOS level) and it is difficult to reverse engineer the registers.
If the RDNA cards have better low-level programming then they won’t be affected.
Note: when I say better I mean that they have proper initialization and reset features. These sort of features are only tested on professional series cards with it being a crapshoot on consumer-grade cards.