Thanks for the explanations, i think i see a bit more clearly now. Since you seem to have a deeper knowledge and understanding about these things, i would like to have your insight about a matter that is roaming in my head for some time: We know that ACE/HWS blocks responsible for ASNYC run up to 22% faster in PS5, couldn't this also contribute to higher compute efficiency of PS5 on top other factors such as lower number of CUs clocker higher and cache architecture, thus reduce the real world compute power difference even further?
PS5 and XSX have the same number of ACEs and HWS blocks. Fundamentally, these blocks just schedule graphics and compute workloads to the shader arrays. I wouldn't say, however, that the PS5 has an advantage due to these scheduling blocks running at higher clocks. Or if it does, I wouldn't expect it to be major. Fundamentally, MS/AMD aren't going to spec the XSX GPU with a number of ACEs/HWSs insufficient for the overall GPU design. I suspect in which case, PS5 is somewhat over-specced, but it's more likely that the PS5 will find less instances of being bottlenecked by front-end workload scheduling than XSX. Although I don't see workload scheduling being a bottleneck for either console in the prevailing majority of cases.
Was not part of the discussions and some rumblings post launch that the Enhanced GE itself is programmable and developer can alter its behaviour but need to code explicitly for it (or take it in its automated default form) on top of the shader cores based geometry culling step?
My understanding is that the customised GE in PS5 allows for fully programmable Geometry Processing steps in the rendering pipeline. So traditional vertex shading, hull and domain shading steps, can be computed on the shader cores (much faster and in a much more flexible way) using Primitive/Surface Shaders. Culling is included in this.
So the "fully programmable" part of it allows you to write custom PS/SS to do things that simply weren't possible previously, but as you say it needs to be targeted specifically by the developer. All the way back from Vega, it was implemented as a separate data path (AMD calls it the NGG Fast Path) independent from the traditional fixed function render path. NVidia on the other hand, replaces the existing geometry pipeline entirely with the new geometry processing pipeline with Mesh/Task Shaders. So both companies approach the problem in two different ways.
Dreams did release in 2020... it had been out for a month and a half before their FY2020 end of March results. And their revenue was terrible; a massive step down from years past.
That's sign of a major flop.
Dreams was out in open beta form to every rather a bit before 2020. 2020 was a soft launch.
Their revenue dropped but if you look at the other number in their filings it's clear there some other things going on. Fundamentally, you're right their sales revenues were down, but a big chunk of MM's recurring revenue came from LBP and has done for a good many years.
I'm also confused by the numbers you are presenting; their records show an operating loss of ~200k for FY2020, and an operating profit of just over 1 million for 2019. They are a company in decline, who launched a game a month and a half before the end of their last financial year, and managed to lose money.. and have 8 million less revenue than the year before.
1m was only in 2019. Almost every year prior to that going back half a decade their Op profits were around 3.5 to 4m GBP.
Yes their revenue was in decline because their legacy titles were drawing down on their natural sales life-cycles.
Fundamentally, they've been a profitable studio and that's what is important. They only need one new solid title to put them back in the green, and because they're such a small studio with low operating costs, there's no reason to think Sony would want to shut them down. They're extremely talented studio.
