A Hacker News thread discusses Mark Cerny's involvement in AMD's RDNA architecture, specifically noting his contributions to "big chunks" of what might be RDNA 5 or a similar custom iteration. While AMD's roadmap shows UDNA as the next major architecture, some speculate Sony may have requested a custom RDNA variant. Discussion also touches on the slower pace of generational leaps in console graphics, blaming bloated game engines and a focus on incremental improvements rather than optimization. Some users point out the dominance of Unreal Engine could contribute, with its default rendering pipeline not always being the most efficient for all games. Other arguments revolve around art asset size and the increasing reliance on post-processing techniques like TAA. Additionally, the discussion covers backward compatibility, high expectations, and potential GPU improvements with the rise of the current APU market and its effects on cross platform engine decisions.