Unreal Engine 5 Tech Demo, [UE5 Developer Availability 2022-04-05]

need both, well technically not RT, but you need a better lighting/shadows solution (that UE5 provides) to help bring out the detail in the dense Nanite meshes, or you won't really be seeing the difference.

 

You're in luck because Nanite is still triangles and polys just a lot more of them. It's still rasterization too just running on shaders and not the hardware rasterizer. Nanite's secret sauce is intelligent compression, streaming and culling of these massive meshes before they're sent off to be rasterized.

Lumen is where things get really funky though.

 

Thanks guys. What I just can’t figure out is why I’m still playing games with visible poly edges, yet somehow this one bit of tech is pushing gazillions polys around like it was nothing. It’s a humongous jump.

 

Even without things like Nanites, it will still take years for the baseline geometry detail to increase lot due to old consoles, weaker PCs etc. Publishers generally move very slowly to maximize profits over technology.

 

It's kind of fun that it's not actually pushing more than around 20m polys. They just worked a way of having a continuous level of detail, so it all stays at about one triangle per pixel.

 

Trees and grass still look like arse.

 

I wish I could say the same but for me we are still at the stage where lighting is just distractingly and obviously wrong. I will agree with you when the problem is the quality of the effects rather than them just not being there.

When we are past the point where objects are just straight up missing shadows, and mirrors have totally fake reflections then density of the environment will e the number one priority.

Luckily UE5 is tackling both issues.

That's not true though, Lumen is a RT solution specifically made to bring out the high Geo environments.

 

To make complex geometry look good, you need good lighting - as what is driving that geometry are materials. Light is more important than geo fidelity or texture res, by far.

 

I remember when DX11 first came out with tessellation, I was so excited at the thought if insane polygon budgets and as the years rolled by and the geometry didn't improve my dreams were slowly crushed.

 

Although I would argue they're not as noticeable as a huge increase in geometry.

Look at Quake 2 RTX, sure it's lighting is state of the art but over all it still looks like ass because it's still Quake 2.

 

They are fundamentally incompatible the way NVIDIA does raytracing for the moment.

As I've said before, I wonder if NVIDIA would have spend more time on adaptive LOD if Epic had delivered on Sweeney's promise of a compute based rendering engine a little sooner. Now Nanite comes at the perfect time to show the problems with RTX.

 

That's because Quake 2 RTX doesn't/does have a PBR pipeline as complex as other efforts in this space. ie for a game that looks entirely made of metal, there doesn't seem to be any metallic surfaces/textures that I can see on any gun or level.

So light transport in that game is still relatively dull despite being lit by ray tracing.

Minecraft RTX which is just cubes, and you can't get more basic than that, has a PBR pipeline with great PBR textures which is why it looks so damn good for just blocks.

 

Even giving it a complex PBR pipeline wouldn't help it.

The game will always look like ass and dated because of the PS1 era polycounts.

 

I disagree. It's textures are terrible. They decided their best to stick with the original quake 2. A set of brand new PBR textures for the entire game would make a dramatic impact.

 

I agree in the abstract but don't think this is so simple. A cheaper voxel or surfel gi system can roughly approximate good gi, whereas no matter how many rays you cast you're not going to be able to reflect and shadow detail that's only in a normal map. Lumen and the excellent shadow maps paired with nanite make a solid case that we need more geo to really render interesting scenes.

 

We've already had the Geo vs Lighting war more than once.

 

Textures still won't help it as the base geometry would still be insanely low and it would just be looking like a 20+ year old game with a texture pack.

 

There is nothing incompatible with DXR.
They use geo proxies and they just need to add automatic geometry merging for scenes with kitbashing, the same stuff they do for per object SDFs merging into global SDF (probably slightly more complex since we are talking about polygons rather than voxels)
Geometry simplification and merging has been done for tens of years for RT lighting baking, etc, so that is a well studied problem.

 

Nah, it's because it has a maximum polygon count of 4096 triangles per model.
You can only go so far when polishing a very blocky turd.

 

minecraft looks great and it's a literal block.