You're trying way to hard to critique engine design decisions when you have zero knowledge of the games graphical pipeline and haven't seen one lick of code. I'm pretty sure sucker punch's team deliberated over every design decision and collectively agreed on targets in response to their game design requirements. You simply pointing at cryteks path and repeating hey just do what they did is as flawed a concept I've ever seen.
First of all check my first post in this thread.
Second of all if Crytek is doing same render target in lower precision without losing quality that means that they spend less bandwidth with the same effect, so their solution is more efficient by definition.
And finally The Order is using 4xRGBA8 not sure about one, we have only one inconclusive slide about that [in material properties section].
I dont know about Killzone SF, because in post-mortem demo slides they were still finding out what is the best precision setup, but they used 5x buffers + 32b stencil/depth, but even this unoptimized setup is smaller than Infamous one.
I'm using Crytek data, because they gave the most inside into development from both Crysis 3 and Ryse and they actually researched g-buffer packing methods and their efficiency and quality in the past.
In short, there is more gain to be had even at this point in time.
Yes, exactly. Maybe not for Infamous SS, because maybe they are not BW bound in any situation, so changing framebuffer setup would not change much, but in future sure.
But maybe that setup wouldn't let them do something that they otherwise can do. I'd be very surprised if these guys were not aware of the years old Crysis 3 buffer breakdown. Also that saving seems like a drop in the ocean considering that they are using 290MB for render targets so the bottleneck might be somewhere else entirely anyway.
As i said in this post, they could not be BW bound currently so the buffer setup is not bottlenecking them, but it is inefficient.