"Unlimited Detail" Tech still being developed

[FordGT90Concept]

Let's take a step back and look at it from the macro game design perspective:

Benefits
-Substantially cuts down environment creation time because its foundationally materials instead of art.
-Does not require multiple models of varying levels of detail.
-Does not have an arbitrary polygon budget to stay with to keep framerates acceptable.

Costs:
-Substantially more hardware resources required.
-Does not look as good because artists can't give fine tune assets.
-Lighting is hugely problematical and, in night-time environments, could lead to massive bugs where areas are inadequately lit.
-No one presently has hardware designed expressly for the purpose of accelerating cloud point graphics where OGL/Vulkan/D3D hugely accelerates polygons.
-File/download size of the point cloud assets are massive.
-Animations are problematical.

I don't think this is viable, at all, until:
1) they get hardware vendors to support it. I think the only way that would happen is if Microsoft buys out Euclideon and adds it to D3D spec.
2) the problem of lighting is solved.
3) the technology advances to a point where, at face value, the visuals compete with or beat polygon renderings.


The film industry has dabbled in point clouds but...yeah...not impressed:

 

[silentbogo]

Let's take a step back and look at it from the macro game design perspective:

Benefits
-Substantially cuts down environment creation time because its foundationally materials instead of art.
-Does not require multiple models of varying levels of detail.
-Does not have an arbitrary polygon budget to stay with to keep framerates acceptable.

Costs:
-Substantially more hardware resources required.
-Does not look as good because artists can't give fine tune assets.
-Lighting is hugely problematical and, in night-time environments, could lead to massive bugs where areas are inadequately lit.
-No one presently has hardware designed expressly for the purpose of accelerating cloud point graphics where OGL/Vulkan/D3D hugely accelerates polygons.
-File/download size of the point cloud assets are massive.
-Animations are problematical.

I don't think this is viable, at all, until:
1) they get hardware vendors to support it. I think the only way that would happen is if Microsoft buys out Euclideon and adds it to D3D spec.
2) the problem of lighting is solved.
3) the technology advances to a point where, at face value, the visuals compete with or beat polygon renderings.


The film industry has dabbled in point clouds but...yeah...not impressed:

It is much simpler than that.
All we need, is for MS to hire John Carmack, and by the same time in 2022 we will have SVO Raytracing as a part of DX13 API

You do realize their company only has 35 employees, right?

Here's a quick demo of a voxel terrain engine, made by 1 person:

You do realize that 35 people and 8 years of time add up to a lo-o-o-o-ot of man-hours.

 

[Nuckles56]

Just looking at how long it has been in development for so little to show should tell you that there is no way this is legitimate

 

[FordGT90Concept]

Voxels are still fundamentally polygon-based. From that video pasted above, it sounds like they're using the CPU to pull points of data from the cloud and generate every frame from them. There are no polygons, just pixels, what's more, you could remove points (like a voxel, but much, much smaller) and it will show those changes in real time.

I could really see this technology taking off if a card were made with a co-processor designed specifically to pull points from a cloud and render it. The most important feature of it would be a massive pool of RAM. If you could load the entire cloud into a cards memory so it no longer has to bother the CPU except for changes, I think it's doable. As RAM density grows, this sort of technology becomes more feasible.

Like I said, Eucludian needs help from hardware manufacturers.

 

[Steevo]

Voxels are still fundamentally polygon-based. From that video pasted above, it sounds like they're using the CPU to pull points of data from the cloud and generate every frame from them. There are no polygons, just pixels, what's more, you could remove points (like a voxel, but much, much smaller) and it will show those changes in real time.

I could really see this technology taking off if a card were made with a co-processor designed specifically to pull points from a cloud and render it. The most important feature of it would be a massive pool of RAM. If you could load the entire cloud into a cards memory so it no longer has to bother the CPU except for changes, I think it's doable. As RAM density grows, this sort of technology becomes more feasible.

Like I said, Eucludian needs help from hardware manufacturers.

Considering people can get custom silicon developed for mining and all sorts of other stuff, they need more than hardware support, they need to get a clue.

 

[FordGT90Concept]

Based on existing instruction sets. Cloud point systems need their own sets. Most consumers of cloud point systems right now don't have a problem spending $4k+ on cloud point systems. That's not something in reach of the average gamer.

 

[qubit]

Considering people can get custom silicon developed for mining and all sorts of other stuff, they need more than hardware support, they need to get a clue.

Love it.