If the author there makes the point to say that it is a labour intensive faked system for GI that only works in specific scenes - what is the purpose of pointing it out as being an alternative or relevant performance metric?
I think for a balanced discussion it's important to showcase as much as possible before dismissing any of it off.
That being said, there is a stark difference between a demo like the ones posted, and full implementation into a complex AAA title.
Of which DXR has been successfully bolted-onto two titles so far, with a couple more expected later this year.
I don't know where these compute solutions will go or whether they will surface into anythign more significant, but it's good to keep tabs on what can be done outside the realm of RTX.
But obviously I think it's fair to say that we're not comparing this tech demo and it's performance to the performance of what we see from Nvidia RTX cards and BFV and Metro. I mean, lets face it, metro and bfv without RT is going to draw away tons of available graphics performance already; it's questionable how much performance is left on these cards for RT to be done and thus why I still think hardware accelerated RT is the way to go right now, at least for a generation or two.
At the core of the debate there is performance and there is labour cost. The second usually gets omitted in the discussion. But I am a believer that the second is certainly as big of a driver if not larger than the first.