Sure, again just my theory. I understand that its not as simple as choosing a GPU w/ the same number of compute units. Still work to do to make sure they behave same as PS4. My assumption is that the decision was made "quickly" to go with a chip w/ 36 CUs and variable clock so that they could create a 1:1 PS4 Pro profile. As for the 40 CUs, yes I also understand that is done so they can manufacture chips without as many defects. When I say "later on" I still mean while the hardware was on paper.
What I am getting at here is that my guess is the choice for the GPU was made to more "easily" maintain parity with PS4 in order to achieve backwards compatibility. I think over the past few years at Sony there have been two prevalent strategies that really clashed: one focused on recreating the success of the PS4 with exclusive single player titles in their own tightly managed walled off ecosystem and another strategy focused on positioning PS as platform with more of a focus on PC and PSNow. Seems like Ryan represents the latter and that is the strategy that won out. I am not sure how focused a 2017 or 2016 Sony would have been on BC. I think it came as later development (relative to the hardware dev process) where they looked at all of the GaaS/Live titles their partners had on the system (as well as the digital revenue they brought in) like fortnite, Destiny, FIFA, Minecraft, GTA, etc and realized that they needed to support those titles crossgen. I could be totally wrong but thats just my read on the leadership shake up over there along with some of the changing messaging we've seen over past 2 years. It's why I think they chose the route they did in order to ensure they achieved back compat as tightly as possible.
No, your theory still doesn't make sense. The problem is a lack of information, where people think that certain processes or directions can be quickly changed. These next generation machines have been in the works for at the very, very least 3 years, if not 4, they have a chat with the vendor, AMD in this case, and enter into an agreement for the development of a new machine. These are after all really big projects, as the components inside of the machine are customised quite a bit and usually they are the biggest APUs AMD has ever produced, so it's not a simple task. Jim Ryan has only been the president and CEO of Sony since April 2019. Your timeline doesn't make sense, as by the time that happened, the design was mostly locked in.
Your theory doesn't make sense because BC is not THAT complex to achieve when the previous machine is based on the x86 architecture, as well as the GPU coming from an architecture that the new generation has built on heavily. Not to say it's anything close to simple, but it's not like it is trying to emulate an ARM or a PowerPC CPU. You can create a 1:1 PS4 Pro profile through software (or firmware? I'm a Java dev, I don't do such low level code.
), by turning off the required amount of CUs to have only 36 active, that's not a good explanation. Aside from that, regardless of the strategy, you think Sony would go through all the bother of creating an ecosystem and then throw it away for next gen for no good reason? The eco system is very useful for them as it helps them to lock in customers and transition them easier to their next gen hardware, as for example Apple will gladly explain to you.
Sony isn't "focusing" on PC, if that were the case every first party title would be available on it. The real explanation is that Sony is probably struggling a bit to grow the PlayStation userbase even further, so it's probably hoping that by porting a few old titles to PC, they can let other potential customers of the type of games they have been missing out on. Regardless, even if your theory was correct and they want to focus on that, that doesn't preclude them from creating the most powerful machine they can within their constraints.
Now, having explored those reasons you gave, I'll give you more sensible ones. As described in the presentation, you can go narrower and faster, or slower and wider. What is the biggest difference there? Cost, cost, cost. As long as you don't need exotic cooling, Sony probably went to that route since a smaller number of CUs have higher yields and are cheaper to manufacture compared to Xbox's 52, and decided to go with a rather innovative way of increasing the clock speed and get the most performance out of the existing CUs as possible. From what I understand, it makes sense as well, as different tasks happen at different times and there are times where there is more of a load on the CPU and there are times where there is much more load on the GPU, so by shifting power between both components dynamically in a deterministic way, you can be very efficient and extract more performance than you normally would out of that chip.
You also have to remember that they had to take cost into account due to the I/O sub system, since that fancy SSD isn't cheap. And in the future, they'll be able to more easily create a slim and/or a Pro version of the console by taking advantage of fabrication node improvements.
I know that it's more exiting to think of drama in the office and game of thrones like betrayals, but it doesn't work like that.
www.theverge.com
blog.playstation.com
www.eurogamer.net
www.playstation.com
www.tweaktown.com
