drotahorror
Member
You know when there's voxels involved, there's gonna be some cool physics. W/e this is, it looks sick.
I'm speculating based on paper specs. It should be possible to keep the same level of physics simulation, as the XSS CPU still looks very powerful. Maybe graphics settings would need to be lowered, but I still have a lot of optimization work planned. My game doesn't use much memory (no textures, no open world streaming yet, instanced meshes), so I should be able to stick to the faster subset and the memory bandwidth there looks ok (8GB @ 224 GB/s).
EmberGen is really cool. I think we will be seeing it pop up in various games. There might not be volumetric export yet but people can still export image sequences. On their site they say Bluepoint is using their tools.
I don't know too much about the memory but yeah I imagine OS and any game data that's not updated/read as frequently? Hopefully the CPU also gets access to the fast memory because I read online that on the X the fastest 560 GBps is reserved for the GPU and the CPU can use the 336 GBps memory.Why are there two different blocks of memory for the series s? The smaller one of 56 GB a second what's that one for can it be used for the video games at all or is it just for the operating system?
Now hold on. Your demos have been mind-blowing to me are you saying that the series s could play video games with that level of destruction and physics? Are you saying that games on the series s are going to have such unprecedented levels of destruction like that?
Yep, rtx are more impressive than physics...said no one ever...
Complex physics, especially particle physics should never be run on the cpu though..... The gpu does that 10-20 times faster due to its parallel processing nature. CPUs reach their limits in simulations like this very fast.I'm speculating based on paper specs. It should be possible to keep the same level of physics simulation, as the XSS CPU still looks very powerful.
If the GPU is just doing physics, definitely, but if it has other rendering tasks it needs to handle, CPUs these days are not that bad either. These 10x-100x comparisons often came from comparing GPU to non-parallel, non SIMD-vectorized code. AVX2 (8-wide floating point) is now fairly common and AVX512 (16-wide) is starting to be as well. Core count is also increasing.Complex physics, especially particle physics should never be run on the cpu though..... The gpu does that 10-20 times faster due to its parallel processing nature. CPUs reach their limits in simulations like this very fast.
Initially focusing on DirectX 12 supporting systems and once that is all stable I might branch out. I'm just one guy and if I spread myself too thin all versions will be mediocre.Is a PS5 version planned?
Still no comparison to gpu speed here......almost as if their architecture was made for exactly that type of mathIf the GPU is just doing physics, definitely, but if it has other rendering tasks it needs to handle, CPUs these days are not that bad either. These 10x-100x comparisons often came from comparing GPU to non-parallel, non SIMD-vectorized code. AVX2 (8-wide floating point) is now fairly common and AVX512 (16-wide) is starting to be as well. Core count is also increasing.
What happens when Lockhart is announced will the game be able to be played there? Because if it can he needs to stop acting like it's not possible on the PS5.
You are totally right, that's why something more than physics in Control is going to run on GPU. Which is bitch on itself, because you have to sync those threads and that's not an easy task.Complex physics, especially particle physics should never be run on the cpu though..... The gpu does that 10-20 times faster due to its parallel processing nature. CPUs reach their limits in simulations like this very fast.
I miss those