Unlimited Detail

Hello guys & gals,

I have a suggestion that could possibly change the way that you game with “Unlimited Detail”. I was searching through the internet on how to make things look better and more realistic in a video game and I came across something really interesting that I think would benefit Epic. Here is a video . The way they do this is, Instead of using Polygons they use “Atoms” which they stream straight from TB hdd’s or the internet. It would b e neat to see Epic adopt this to make awesome looking games. There are also more videos explaining how this is possible.

Is this possible to implement?

This is just point cloud data and their ‘tech’ is different way of loading the data from disk.
Ppl have discussed about this to death for many years and the thing is: it’s not revolutionary tech and is quite old and also isn’t suitable for game development. They just put out fancy names on things and make bold statements, but in fact that ain’t going anywhere for graphics in games. You can simply “remove” polygons from game technology, won’t happen.

Also, that wouldn’t create unlimited detail, you still have to make everything you want to have in there

In that video you’re looking at results of several years of research, so, unless epic decides to collaborate with euclidian, it is not gonna happen.

Amount of data required by those scenes is massive. If you listened to the video you’ would’ve noticed that the narrator casually mentioned that the data is streamed from multi-terabyte hard drives.

Being able to stream and display this amount of data in real time is actually a big deal.

It is really disappointing to see this kind of statement on game engine forum, especially when that game engine is unreal 4.

The tech is quite obviously useful for VR environments and digitizing real worlds. The obvious application is oculus rift. There are plenty of reasons to want something like that.

It’s not any more of a benefit to VR than the current polygon system is. If you’re doing photogrammetry, then the results using current methods that generate a mesh will have just as much detail but far less data.

Asking the industry to remove polygons from games is like asking car manufacturers to remove gas engine from cars and fuel them with water instead. It’s possible, but nobody will do it in this lifetime because of ‘reasons’.

Photogrammetry-based polygonal model will produce noise at certain distance, in this video I’m not seeing any.

Are you even serious? There are cars on the market right now that do not run on gasoline. Tesla motors went electric route, and asians have hydrogen based vehicles, there are countless other alternative approaches.

The point is, polygonal approach to 3d has severe limitations and makes certain tasks (like CSG) too difficult. You also HAVE to use lods with polygons because when there are multiple triangles visible per pixel, it will produce noise.

This industry is supposed to innovate, you know? That’s why I said that seeind this kind of comments is depressing.

Please don’t bring drama into this like the other hundred times this topic has been brought up.

This technology has been promised a long time ago but all I’ve seen are the videos they post, nothing from a major third-party studio. In any case, if a UE4 engine developer can chime in, he’d be more knowledgeable about anyone here on whether or not this tech is feasible for inclusion.

That said, I know UE4 is used for many applications but gaming is where it started and where it’s still used the most. And I doubt anyone wants to download terabytes of game data just because a tree looks better now that it did before.

We dont need euclodio or what ever. We still havent maxed out our actual technology yet.
The last big innovation was just displacement and raytracing. And we just scratched it.
No way we will get to “atoms” in the next 20 years.

This just seems like all of those “revolutionary” projects on Kickstarter that raise 2 million USD (Razors that last forever, etc.) , and then you never really hear about that product again.

Since when?

Just like oil industry does everything needed to stop threats to their business, GPU manufactures will make sure things like this never take over unless under their financial domains.

And electric cars are still indirectly attached to the oil industry; Tesla Motors are trying to be the ‘Apple of Cars’, but their business aren’t running very healthy right now. Although I know nothing about cars, is easy to see why.
Also, the japanese guy inventor of water based car engine simply vanished, nobody have any idea of him anymore.
If this tech could be that important, NVidia would do anything to buy them out and then destroy the tech they have.

And yeah, I’d say that ‘depressing’ is a mark of reality. The world we live in…

Euclidieon is a very dishonest backer seeking organization with no real innovative tech as far as most are concerned and they tend to hide their limitations and make specious claims. They have some recent progress that seems interesting but their past history alone makes them discountable. If you actually want promising work in the field of sparse voxel octree based environments then the Atomontage engine is what you would want to look at.
This guy actually has promising work and even though some of the claims (600fps on laptop) seem a little suspect, he still seems legit and has some very impressive videos.

That being said there is literally no reason to try and switch rendering paths at the moment, all toolsets and graphics cards are optimized for Polygonal rendering and the basic workflow and tech for fully voxelized scene raycasting/rasterizing isn’t there yet to make a switch over worth it yet/ever. Even atomontage admits that most of his (fairly small) scenes take “several dozens of hours to generate on a single core and tens of gigabytes of data before compression”.

That is not unlimited detail and those “atoms” are pixels. This tech has existed for years and cannot be used for games due to the very obvious limitations.

Since always. When you have very high polygonal density and multiple polygons per pixel, you get noise. That’s one of the reasons why lods and normalmaps exist.

Google Toyota Mirai. Hydrogen, not water, but not gasoline.

NVidia doesn’t have to care. As long as the new tech uses GPUs, it will bring more money to nvidia, no matter what kind of tech that is.

I’m not saying Euclidieon is trustworthy or something. In the video op linked they promised to show a game that uses their tech, and the video was posted 1 year ago, and there’s no game.

However, bashing any new technology just “because it isn’t polygons!” on gamedev forums is ridiculous.

Not much difference compared to lightmap baking or Realtime precomputed gi.

Maybe something like any kind of voxelization using any unknown photonics properties and electromagnetism.
This way the details will be volumetrically replicated at subatomic level
Then polygons will no longer be necessary, no more.



They have been posting videos promising the world since 2009 or so and changed their name from Unlimited Detail to Euclideon. And yes there is a massive difference between lightmap baking for a world map taking a long time and some memory and 30+ gb of data and over a day of processing for a 100’ x 100’ scene. I am not bashing anyone for not using polygons, I have been playing with voxels and voxel rendering systems for years (have a plugin for this very engine using them) and do think that it is the way of the future, but that is years off still and when it comes it is very doubtful it will be from Euclideon, that would be why I linked the closest to viable alternative from someone else working on the problem and showing a working prototype.

No, LOD’s exist because if something is only taking up a small number of pixels on the screen then it doesn’t need the full detail. Normal maps exist because it adds small detail that you would have to model otherwise, and because they run faster than bump maps since there’s no depth calculation required.

This company is selling snake oil - it’s nonsense, I’d go so far as to call it a con or a scam.

Their tech demo has barely changed in the 12 years that it has been floating around; no progress has been made and what they have is borderline useless for most applications. For valid applications like medical imaging, there are plenty of other companies who have been doing it better for quite some time.

The irony is that in recent years smaller indie games like Spintires have made much better use of voxels than they’ve managed in a decade.

This (voxel based) tech would be great replacement or alternative for current landscape system. But done statically in editor, so it can get shadows and what not in runtime. Also that stuff needs some way to make it look less minecraftish at close ranges. But as landscape tool this would be really awesome, then add some tools to make procedural erosion etc.

As standalone game tech it can do some rts, flight simulator, anything that shows terrain from some distance, up close, it is just ugly. It is bit too early for detailed voxel games, but in few years (maybe Skyrim 3 or Fallout 5).

And no demo videos do not count, they are just tech demos. For reference see what nvidia does on titanX now, its awesome but we need 2-3 years for games devs to start thinking about that tech seriously, then next 3-4 years for them to develop games. Current games (Fallout 4) look like best tech demos from 2009 or so. Tech needs time to go from limited tech demo to full game, games have ton of other things to calculate besides cool graphics effects.

Anyway this voxel stuff would be awesome for top down rts.