| goopy20 said: I was a bit baffled by Sony's presentation but I have to admit that the SSD tech definitely sounds like a game changer in the way game worlds can be realized and how we move around in them. However, if developers would fully embrace the tech and build their whole games around it, how will that work on Series X and how about pc? |
We do not know.
What people don't seem to understand is what was actually solved by Sony. Whether Sony spent too much money on this (and sound hardware) and sort of forgot the rest (I'm not happy at all about the memory bandwith, for example), only time will tell. Here it gets very technical, but the short version of it is: Sony can stream into gpu memory space with gpu cache coherency, and do this at very high speeds. No ssd/pc combination in the world can do this. Yes, there are very fast ssds available for pcs, but they all stream into cpu memory! The XSX also streams into gpu memory space (not quite at the high speeds as the PS5), but MS has not explained how they solve the problems associated with that. The cache scrubber (and possibly whatever is in the Kraken hardware) is proprietary to Sony, so MS doesn't have that.
I remember a demo of AssCreed where the guy runs through a crowd. Often people and stuff appeared out of nowhere (what is called popup). The faster the guy ran, the more popups, very annoying to watch.
Now on the PS5, these popups are entirely gone if you program your stuff "the correct way". Even when you turn around your warrior and s/he looks elsewhere and a whole new lot of assets have to be loaded - no popups. That situation can easily get completely out of hand when you use ray tracing in building your scenery, as your gpu caches might no longer contain large parts of the correct scenery.
The games will show us the truth. I imagine that a game tailormade for the PS5 will run atrociously on a pc, if it is a simple port to the pc (whatever that means). What the XSX does with such a game is anyone's guess at this time of (not) knowing the intrinsics of the gpu hardware, but I can imagine it will require some additional work to "get it right", a simple compiler flag setting won't do the job.
