Mostly because it is a stretchy foam suit. A model is deformed and translated based on animated bones in the model. Essentially each vertex is tied to a small number of transforms matrices(bones) with multiplier/weight. Animations are updating the transform matrices over time. You only get so many bones per model and per vertex and because of this clothes moved inline with other parts of a model like skin because its using the exact same bone.
I'm not sure where we're at in terms of increasing bone counts or independently animating clothes on a character. As bone counts increase, the amount of matrix data that must be sent to the GPU every frame increases.
Are you suggesting building some kind of GPU side physics protocol that takes material properties, and existing bone data and fills in some of the blanks? I suppose you could start to push bone motion vectors and interpolate some things.
My main thought is just that DL reduces the amount of computate you actually need by allowing you to approximate rather than derive things like cloth physics, how that's eventually integrated seems an implementation detail.
Maybe once real time photorealism is cracked, we'll move to a completely Newtonian global physics for some games, with some analogue of "atoms" the way RTX models "photons."
I'm not in the industry, this is just something I'm wondering about. What would Minecraft be like if blocks were the size of pixels and the physics were real?
There is an extreme amount of doubt about whether the tech is real (myself included). These people have promised the world, have incredible videos, but have nothing to show in terms of a demo you can download and run yourself. Still, they claim to do what you describe and have videos demonstrating it: https://www.youtube.com/watch?v=J62z_7JaYMw
That type of detail suggests to me that it's a real (clunky) demo, but maybe it's running slower than advertised.
I can also understand not releasing a demo.. the code would be too valuable to risk reverse engineers.
Also this demo is supposed to be running on a PS5 devkit, which means that you'd need a devkit to run it, which means that you'd need to sign NDAs and join their developer programs and whatnot.
Having worked with current gen consoles (meaning I can't go into any amount of detail), it's not a trivial thing to get a demo like this running well on PCs. This demo is likely making use of every platform specific feature available to them.
That said, the demo might be accessible through some back channels if you're already a UE4 licensee and have a PS5 devkit.
> it's not a trivial thing to get a demo like this running well on PCs
This old tired argument.
The current theory for how the meshing is done is something similar to mesh shaders (available on commodity PC hardware since 2018)[1]. This "PS5 platform specific feature" running on PC in 2018[2].
As for the lighting, NVDIA have already had this "platform specific feature" on PCs for some time now. It's called RTX. In 2018[3] (using DLSS), in 2020[4] (no apparent DLSS usage, but it may have improved).
Both next-gen consoles are essentially PCs. Their primary advantage is tightly coupled hardware (e.g. memory latency, the absurdly fast PS5 SSD). While dedicated raytracing silicon on AMD is currently unique to PS5 (AMD claims they can emulate DXRT on Navi), it has been around for more than a year in consumer hands in the form RTX.
[1]: https://devblogs.nvidia.com/introduction-turing-mesh-shaders... [2]: https://www.youtube.com/watch?v=CRfZYJ_sk5E [3]: https://www.youtube.com/watch?v=jkhBlmKtEAk [4]: https://www.youtube.com/watch?v=2744rWPvNuE
On the CPU side yeah it's 100% just a normal computer but nothing will be interrupting your threads. I think Windows 10 tries to do in it's new game mode too.
Sorry for assuming the link was the PS5 one. I have a UDN account and their login setup sometimes just dumps me to their homepage, so I made the assumption that it was the same video that I had seen everywhere else.
Indeed it didn't, and DX12 is extremely old tech. Consoles don't have async compute right now. (Edit: apart from the PS3).
The PS3's SPU definitely counts as async compute especially with how it was used later in the console lifecycle[2] once people had time to familiarize themselves with it.
However, in the current gen consoles, you don't have to deal with a different ISA, command queuing, and shared memory between the GPU and CELL processor. You are only writing HLSL/GLSL/PSSL and setting up an aggressive amount of fencing to transition resources between readable and writable states within the GPU.
[1]: https://www.anandtech.com/show/9124/amd-dives-deep-on-asynch... [2]: https://www.slideshare.net/DICEStudio/spubased-deferred-shad...
There are a few indie devs trying to make "finer resolution" Minecraft; e.g. https://www.youtube.com/watch?v=VQv1OEm_www
There is no way we can simulate physics at the level you suggest for real-time games, not even in a hundred years. Unless you achieve a breakthrough in fundamental physics which would have an impact in every aspect of life, not just games.
There must be some granularity at which voxel engines can work smoothly today. Let's say my room is a square with 320 pixels on each side. 320^3 = 32,768,000 voxels, which looks like a lot, but an RTX 2080 can render 17 billion triangles per second, which works out to 273 million per frame at 60fps.
Let's ignore gravity and focus on momentum vectors and collisions. Even if simulating physics for each voxel were 10x slower than rendering a triangle, you could do 320^3 voxels in 60fps on today's hardware.
And this is naive, assuming you can't merge neighboring voxels together! I think you're being unfairly pessimistic.
Current systems simulate about 30k simple bodies. So that is already off by 3 orders of magnitude.
And that is a basic simulation, nothing complex, no fancy physics simulated, just rigid collisions.
Even if you could simulate the numbers you propose, 320 voxels per side is really low. Each voxel would be around a centimeter wide, which would look very bad (compared to the usual way of doing things).
...that now it's the actual quality of the models themselves, and even more of character motion, that are going to have to be the next steps in better graphics.
I'm stunned by the lighting and detail, absolutely stunned.
But at the same time, especially when you get into the architectural rooms, everything's far too "perfect". No grime, no crumbling edges, the motion of the ceiling opening is far too smooth, no vibration or catching, etc.
Likewise, the character's ponytail isn't bobbing like real hair at all. Sometimes her movements seem to be anti-gravity and bizarrely robotic. The leather over her shoulder seems to float rather than rest, and twists as if it were made of gelatin rather than stiff cowskin.
Not that these are criticisms at all -- I'm well aware of how hard they are. It's just fascinating to me to see, once rendering is basically "good enough", how the flaws in modeling come to the fore.
And so where are those advances going to come in? Are there going to be procedural-generation advances in creating realistic grime and imperfections and irregularities? Or deep-learning advances for realistic bodily movement? And why is it so damn hard to get clothing to actually fall on the body right, instead of always looking like a stretchy foam skinsuit?