Voxel games

[Zarax]

Voxels have been a kind of gaming holy grail for the last 30 years, giving the promise of the same level of interactivity we had in many 2D games but in 3D, yet aside from some popularity in the 1990s before GPUs came to dominate it never left niche status as a technology.
Teardown so far seems to be the most successful game in recent times, however I wonder what technological limits we have in making truly good looking voxel games in 2023.

What do you guys think?

 

[Zarax]

2D Voxel tech in games like Comanche is quite different to what most people refer to as voxels these days.

When people talk about voxels now, they're talking about any 3D grid based world, Minecraft, Teardown etc. The reason it isn't used to build most games, is that it's a really, really expensive way of representing a world. Regular 3d mesh based world are always going to be more efficient and give better results in 99.9% of games.

There's the odd game where the destructibility it is the core of the gameplay experience and warrants all the compromises. Minecraft, very large voxels, huge worlds. Teardown, still pretty large voxels, very small worlds.

There's quite often voxel-based technology in the back-end of AAA games (most navmesh solutions work by voxelising the world), and things like environmental effect systems, fire, smoke.

But the question is, is it inherently expensive or is that due to lack of optimized engines/hardware?
For example, how hard would it be to make an high res version of Comanche/Armored fist VS a prettier looking teardown?

IMHO it would still be a very interesting tech for RTS/Isometric RPGs where there can be lots of benefits on interactivity and they can be made to be relatively inexpensive on the hardware side.

 

[Zarax]

It is inherently expensive. The density of data required for a voxel world is orders of magnitude higher for a voxels vs traditional meshes.

You could, but it would look worse than just building the same scene with traditional meshes. Modern GPU are very, very good at drawing lots of triangles.

Not sure how it would work with RTS, pathfinding would be a nightmare.

Again, with an isometric RPG, yep you could probably do it, as you have quite a small part of the world in view at any one time, but you're having to invest a vast amount of time into this voxel engine so......what? You can knock a whole through a wall? Just not worth it unless your main USP is being able to knock holes in walls!

I'm aware modern GPUs are good at drawing triangles but then how are voxels different in a computational way? I don't know enough, which is why I'm asking. I could accept worse looks if the tradeoff is a truly reactive world, with for example cover degrading or hills being slowly leveled by intensive explosives for example.

How would pathfinding be a nightmare compared to old style RTS? Dark Reign in 1997 used multi level terrain with several ground types on a 2D engine and pathfinding was pretty good, actually good enough that you could give lots of autonomy to units.

Well, the point wouldn't be just knocking holes in the walls, but world interaction. Just look at the good old Ultima Online, you could have almost unlimited freedom in creating your own corner of the world, something that very few games managed to replicate today.

 

[Zarax]

Data density is the problem. Voxels are not an efficient way of describing a 3D world in a way that GPUs can then render. Not to mention serializing the data for persistence or multiplayer.

A solvable problem, but tricky. Not sure how you handle slopes in a voxel world in a way that doesn't look terrible. There is an immense difference in complexity pathfinding in a static world, versus a deformable one. Going from 2D to 3D.....well, it's literally orders of magnitude more complex.

It's all a trade-off. You make massive technical compromises and give the vast majority of your engineering budget over to trying to make a voxel engine work.....(assuming you can).....so people can interact with things? There are lots of ways to make world interactible without having to go down that road.

So basically what you're saying is that Voxels nowadays are a technological rabbit hole with limited application?
I always wondered why it remained so niche and if it was a cheaper alternative to make an interactive world VS modern engines.

 

[Zarax]

I always appreciate your input on these threads considering your background. I made the mistake when I was younger thinking I'd enjoy making games as I enjoyed playing them, I was very wrong

I know the feeling very well, computer science gave me several kicks in the nether regions when I tried at university

 

[Zarax]

Kinda....it's not a total dead end. The concept of dividing the world into a 3d grid is the core of it....it's quite common still. No Man's Sky terrain is voxel based although it uses a more sophisticated algorithm for rendering than the simple boxes the likes of minecraft uses.

I wrote a 3d navigation system for unreal a few years back that's based on a voxel octree representation of the world.

The crux of it is that fully deformable 3d voxel worlds are very expensive compared to regular terrain + prop type worlds. That may be desirable if the deformability is the main USP of your game, (Teardown), but not worth the compromises if your focus is elsewhere.

I appreciate that it's a very expensive approach but can't GPUs help at all in this? Old voxels used to be CPU based and gave decent GFX for the time, I wonder if with a proper engine and some acceleration we could at least get 2010-2015 detail, like for example a voxel based Red Faction or Supreme Commander.

 

[Zarax]

I'll caveat this by saying that I am an AI programmer and engine/physics stuff is not my area of expertise, but....:

You typically have several different representations of the world in a game. You've got a game object world that describes where all the entities are and what they consist of, this lives in system memory. You've got a physics world which contains all the collisions meshes, also system memory, and you have the rendering world, which lives in GPU memory (there'll be other stuff like audio etc). You update the game world and physics world in parallel on the CPU (which various scheduling shenanigans to make sure things are calculated in sensible orders), and then once you've calculated where everything should be drawn, you fire a load of render commands at the GPU so it can update it's state and then draw the frame.

These days you do have compute access to GPUs, but it's not really that straightforward on PC to take say, the physics world, and farm that data off to the GPU to let it do some of that work, and then pull the data back into system memory so the game logic can do physics queries etc. It's not really straightforward. Not to say it's not do-able....


I think it's more a case that no-one (so far) is willing to risk the millions of pounds required to fund a project to create a voxel-based FPS.

The thought of handling multiplayer in a game like Teardown brings me out in a cold sweat though tbh, managing the data to represent the world at a good framerate on a local single player game is challenging enough, without then trying to stream all that state back and forth through a network pipeline that is the digital equivalent of a soggy paper straw

Yes, I suspect that the amount of data involved in an high detail voxel world must be pretty hefty, so it's better relegated to single player.
Oh well, maybe someday Microsoft will shell the money for a Minecraft 2 with prettier GFX...

 

[Zarax]

Yeah he had some good videos but the tech was obviously a long long way from bringing something to games.

Imagine adding voxels to Dwarf Fortress... It would require more or less the same resources as training another GPT version but you'd basically get the ultimate game in terms of interactivity.