AMD just develop what the vendors want, if Sony or ms want to push raytracing in next gen, AMD will push RT. They make the most money via consoles so they'll cater to them and offload r&d cost between them, why spend 1bln in r&d when you can offload those to your clients?
An architecture just doesn’t happen overnight because you want something. AMD also has interest to a drag console APUs in a direction that also benefit them for PC and professionals/servers. If Mr Cerny wants a full custom chip like old days, it’ll cost a pretty penny. We won’t see those anymore.
Even if AMD went full RT tomorrow, there’s two generations to catch up, well 3 if you consider that they would have to catch up to Nvidia 5000 series.
Just won’t happen easily. They have to ditch the pipeline and start from scratch.
Intel just didn’t land on good RT and ML by luck, they had RT on Larrabee in 2009, although tech as moved way beyond those early attempts, they’re not total strangers to the problem.
What blows my mind is AMD participating in the DXR consortium with Microsoft and Nvidia, looked at early Nvidia talks about where it should be heading, AMD agreed to all of it, as DXR is basically Nvidia’s homework, they saw Turing’s solution, and somehow they managed to fumble around all this time. It’s like the student at the back of the class who was sleeping during the whole exam. There’s no excuses. Intel came late, 10 minutes before end of exam and nailed it.