curl-6 said:
Chrkeller said:
Could be memory bandwidth as well. Either way, we have swayed from the original bump. The bump was someone claiming drive speeds have a significant impact on game performance, that isn't true. Drive speeds reduce loading and are faster at data transfer. Very limited impact to actual game performance. A few PC games have optional ultra texture packs where drive speed can start having a marginal impact, but those packs aren't on consoles, certainly not the S2. Side note, the drives I put in my rig are 12,400 mb/s. I have tossed a 600 mb/s bulk storage drive as well. I can play games from either drive. Outside loading and Rift... there is no difference. |
I still think a CPU bottleneck rather than bandwidth would be more likely when talking about NPC counts as the main demands there are skinning, animation, routines, etc. Same reason we see similar NPC and traffic density in Cyberpunk between Switch 2 and PS4 despite elements like data streaming, textures, etc being much better on Switch 2. Yeah I'm no expert so I honestly have no idea whether faster I/O has any bearing on in-game performance, but I would theorise there's probably some degree of performance boost gained from the FDE, since it offloads decompression duties from the CPU. |
My gut tells me the NPC density is bandwidth more than CPU. The ps4 is 8 gb at 176 gb/s, while the S2 is 12 gb at 102 gb/s. The bandwidth is just going to limit density, fps and resolution. Luckily DLSS is an easy workaround for resolution.
I am far from an expert, and couldn't lay out details, my positions come from experience. I have tried different GPUs, CPUs, Ram, overclocking, etc on my rig. GPU changes immediately make massive impact. CPU makes a solid impact. Ram overclocking was marginal, the other stuff (like storage speeds) was negligible.
Kind of arbitrary numbers but GPU feels like 60% of the game, while CPU is 30%, Ram is 9% and everything else is 1%. I still maintain raw GPU power rules the day. The difference between my old 4070 and 4090 (everything else was identical) was generational, despite both GPUs being the same "generation." It is why I don't think time is a good benchmark for generation definitions, power is. A 3090 outperforms a 5060ti, despite the 3090 being a 2020 GPU while the 5060ti is 2025. Raw power is king at this point in time. Perhaps AI will change that in the next few years.