Funny thing is, that there are also more NPCs in the xbox version in the comparison shots of the DF video. Might just be random, but it is very consistent throughout the video.
But this again speaks for a API problem on xbox side. The xbox one had the stronger CPU but most of the time, even in CPU limited scenes, the xbox one ran worse than the ps4 version. Seems like Sonys API has still a CPU overhead advantage.
Well, that isn't really too surprising considering they aren't writing a devkit API suite that has to accommodate for a wide range of hardware configurations. There's PS5 and that's pretty much it, though it might be arguable that the PS4 falls into that bracket too since PS5's suite is essentially PS4's with a lot of newer things added on top (according to a few devs).
Somewhat related but, I was watching a podcast the other day and one of the guests there (software dev) made an interesting point insofar as "coding to the metal". I think sometimes I see that as an advantage because it implies getting more out of the hardware, but if you really think about it, isn't it kind of funny how "to the metal" programming is now seen as a favorable bullet point? Coding "to the metal" didn't help out consoles like the Saturn, and it made programming on PS2 and PS3 a pain in the ass.
Ideally you'd want to spend as much time away from going into assembly for coding as possible which means you'd want your high-level language tools to be good enough at extracting as much performance from the hardware as possible. That's obviously something Microsoft has been pushing towards for years now (even decades), but I think for some people who get caught up in "teh wahrz" stuff, they don't see how Sony are going a similar route. There's a reason Cerny stresses Time to Triangle so much; they may give the option for low-level programming but ideally they know devs want to stay away from doing that because it just increases development time and leaves less time to focus on the more artistic/creative parts of the game design.
That said I guess dipping into assembly (or as it's called these days hand-written assembly, which at least still sounds like some type of abstraction from raw assembly) is not as much a counter to timely dev if the underlying architecture is well-built and documented. The documentation should ideally relate how parts of each component in the design impact (positively or negatively) performance of other components in the design, but I understand that level of documentation wasn't really a thing back in the older days, partly because so many components of the system design would come from disparate manufacturers. Not that there aren't components in the modern systems that come from outside companies; there are. But they aren't usually the "main" components.
So yeah, it's sometimes odd to see some folks latch onto "coding to the metal" as a perceived advantage in a system's API design because in a way you'd always want the high-level language dev tools to be good enough for any performance you would require. And Sony were arguably the first company to start a wider industry shift away from pure (or near-pure) assembly coding to high-level language support for things like C when they released the PS1. "Time to triangle" was important to them even back then.
