And bleed away all the PS4 profits just like how selling PS3 at a loss bled away all the PS1/PS2 profits? I think Sony would rather not go through that again, just saying...
Not the same situation. PS3 launched at $599 with a $250 loss, 1 year after its competitor, before PS+ existed, and had a bunch of issues with a lack of games and difficult development. This hypothetical PS5 would launch at $499 with a $200 loss (after 14 years of inflation, too), most likely 1 year before its competitor, with a simple and easy design and plenty of games, and PS+ making plenty of revenue.
Yeah, but then they'd just be selling for (effectively) at-cost, instead of at-profit. IIRC PS4 was slightly profitable for Sony from the start, or only sold at a very small lost that was made up with a single game purchase, let alone extra peripherals like controllers, etc.
PS+ might be bringing in lots of profit for them, but not enough to eat $200 on each PS5 they sell, b/c that $200 would actually be a bit more than $200 when you add marketing/promotion/advertising costs, and whatever retailers get off of each unit that's sold (if they in fact do; not sure if that's the case), etc. That'd basically call for PS+ prices to raise noticeably, and that could turn off a lot of people, putting a dent into using that as a subsidizing method.
It's not just about PS+, it's also all the profit they make from games (digital especially), and the most important thing - the head start. Launching a year ahead of Microsoft would be tremendous for them, especially in the market position they're in now, and combined with a huge amount of power and full BC, they might even pull in a bunch of Xbox diehards. The impression i got from the people here is that 2019 would be too early and too weak, so I thought about how far Sony could reasonably push the envelope in terms of power and cost to get it done.
What's dynamic extrapolation? Some form of temporal reconstruction?
Lol, just something I made up, not sure how feasible it would be. Basically like dynamic resolution, except instead of dropping the resolution, you would extrapolate more of the pixels in the way checkerboard rendering does. Checkerboard extrapolates 50%, this would extrapolate different percentages of the pixels as needed - probably 10/20/30/40/50%. Whether it would actually be better than normal dynamic resolution, I can't be sure.