Unreal Engine 5, Nanite and mesh shading technology - when planned for Unity?

Today Unreal Engine published a showcase of their upcoming Unreal Engine 5 where the most impressive feature is something called "Nanite":

According to their words, "polygon budget, normal maps and LODs" - it all will be in the past and no longer required since game engine is capable to render any 3D model with any amount of polygons.

Apparently, their thing which called "Nanite" it is a "Mesh Shading" technology which was shown by Nvidia 1.5 year ago.

That looks extremely impressive and seems like it is an incredible revolution in CG and game development.
If this is true, it may speed up process of optimization and creation of 3D models in 10 times.

However, I am not sure what about game size because high-poly models has a large size, but still, this technology is very impressive.

What do you thing about it?

And of course a question, is Unity planning to implement that any time soon?

 

Well, I found another thread about that. For some reason it was not shown in forum search result initially.
https://forum.unity.com/threads/unreal-engine-5-game-changer.889501/

 

its impesive, they also give infrastructure for multiplayer games for free
to make your games multiplayer easy
I hope so, im devote of unity, im really really hopeful with DOTS in logical performance, but right now unreal is showing the best graphycal performance yet, someday we wil say to our sons "in my times i had to optimize my models to just have no more than 50k poligons" your sons- "WHAT?? Thats imposible"

 

Unity needs to implement mesh shaders into HDRP ASAP.

 

You are funny but this technology seems to be incredible, unity must really integrate a technology like this in their engine.

 

This technology has nothing to do with mesh shaders.

 

Yeah it's more similar to a LOD system

 

anyways i think that it could be super optimized at real time, but the size of the files would become larger than never

 

I'm not convinced of this- larger, yes, but I don't expect drastically so. They will break the meshes into heriarchtical clusters of some sort- most likely at each LOD there will be 16 or 32 entries pointing to sub clusters, until you get down to actual vertices, with some data supplied per cluster level, like average normal and root position as delta's from the parent clusters. This will allow them to quickly find the appropriate data for each pixel like entry they will draw. Since data is stored as a delta from the parent cluster, positions and normals will only need to be stored in low bit depth, making it much smaller. UVs will mostly be continuous, allowing them to compress well as they are effectively gradients. This will likely lead to something like a 20:1 compression ratio vs. a traditional mesh in Unity, and will be able to be progressively loaded into the cache as needed. A compute shader can then transform this data into whatever data the actual rendering needs, most likely id's or lookups into the virtual textures. So there will be a pretty constant fixed cost per frame to compute this data, and likely a minimum size you bother to store data down to. Further, the actual rendering will be a fairly fixed cost as well, as it's decoupled from the actual geometry at this point.

 

Nanite does not seem to be an LOD system. In one of the videos, they showed each and every induvidual triangle being rendered each with a randomized color.

 

Speaking of Unreal Engine, I'm really curious to see, if Unreal Engine will be able to realistically accomplish all of this, for game development.

 

Unity is researching all the time, and I have heard whispers trickle down that there's a fair few interesting optimisations and experiments, but I think Unity is focusing on HDRP and raytracing / the big mmo picture in general first.

UE5 tech is brilliant but it is also not really changing much for gamers. It's changing a lot more how developers work, and may be extremely cost effective.

A seasoned developer can throw more money at a classic engine and still get results gamers believe to be as good visually, and at higher performance - see something like the new rachet and clank. That's insane. That's the sort of thing that isn't possible without where Unity is kind of going with DOTS.

I think Unity should see their path as it is and not chase UE5 in this case.

 

As I'm currently downloading the 98 Gigabyte single-scene sample project that comes with the UE5 Early Access... lol

 

I mean that's source assets, presumably it builds to something more reasonable.

My mobile Unity game has a project size of 10GB that builds to around 600MB.

100GB for a high fidelity scene doesn't seem to be extreme to me.

But to get back on topic:

Unity does not need more unfinished tech that clumsily fits with the rest of Unity.

They need to finish the things they're working on and make sure they fit well together <- The problem is they are currently failing spectacularly at that.

I played with Unreal a bit today, the new UI is nice and the engine feels cohesive. Unity doesn't feel cohesive at all. Each feature is in various states of "not ready" and has its own internal logic, Unity feels like there isn't a plan for how the engine is supposed to work, but there are several teams working in isolation making stuff, and that stuff eventually get merged together with duct tape to create Unity.

 

Fair enough. Granted, I think at some point in the near future, we're going to see them abandon traditional textures and adopt something similar to P-tex. You don't need textures when you've got unlimited polys and if they can figure out how to store the color data in a more compressed format, it might help solve some of the larger file sizes of games today. In fact, you'd probably only need a single color table or something similar and then every asset in your game could just reference the same file.
Oh and the demo is 106.1GB now that it's on disk

 

I agree.

 

So like 5 more SRPs and reboots of input and networking then..

 

I'm not much of a fan of starting something new if the old isn't done. But if Unity wants to try who's to say no

 

I think Unity needs something like Nanite eventually. I would say 99/100 Unreal graphics features updates are not important, just more eyecandy. This is different in that it offers the only real way to render a game in full real-time high poly for the first time. There are definitely countless thousands of games being made that aren't to care about that right now, but for the future I think this is pretty big.

 

UE 5 is a big step ahead and it fixes a lot of issues:

- Nanite: No polygon limit and an no LOD to generate anymore (except for vegetation & characters).

- Virtual Shadow Maps: High shadow density, realistic penumbra and no need for contact shadows anymore.

Lumen is a bit shaky right now. The GI is not physically correct and changing depending on the distance. The Screen space tracing part of it cause light leaking to popin and popout and the voxel lighting part of it cause overshadowing /blob of shadows. However I think in a controlled environment and for game dev it's a great solution.

The cost of this innovation is high hardware requirement but it's made to make game releasing in the years to come.

Unity could catch up 80% of Lumen with ray tracing only (you need other way than ray tracing for some parts of the GI, talking from experience). Also ray tracing is not well integrated in HDRP.

Nanite is the big issue for Unity. For Games, having no LODS and no polygon limits is huge. For the AEC industry, being able to render high resolution cars or data without optimization is also huge. Nanite is years of R&D by the unreal team, I won't expect Unity to have any answer until 2-3 years from now.

 

2-3 years is being optimistic. Given the track record, I am almost certain that it won't happen at all.

 

Gamesfromscratch made a vid on UE5, and the Valley of the Ancients demo game. He goes through a bit on nanite, lumen and other cool looking stuff too. Note, his pc lags a bit while going through the demo, which is understandable, since the demo has alot of stuff in it.

 

It's not ready yet, but they're getting there: Rendering & Visual Effects (unity.com)

And I agree, Unity is able to fight the lumen using only the ray tracing, as long as it is perfectly integrated (well optimized) Only Nanite that I don't see a quick fix coming in now (2021 or 2022). but nothing prevents it from being added eventually, after all, the SRP allows this and is already stabilizing.

 

I don't want any new features until the current ones are optimized and finished, Unity lost its focus a long time ago and it became a complete half-finished mess full of issues, bugs, and very bad QA. It is leaking everywhere, as one member said above, it is holding together with duck tape. As a perfectionist, I would say that even the duct tape is full of holes.

Before I could normally just work on my project, now I'm mostly fixing stuff, figuring out performance regressions, bugs, and other stuff. They presented performant URP as the next big thing with an awesome Boat Attack showcase, even that got massive performance regression that probably isn't fixed till today, unbelievable. I had a good and bad experience with QA but I get a feeling that they expect me to point to a specific line in the code of their engine (which I don't have), even then I should wait 6 months to get it fixed.
Unity is blindly going downhill, I think that they are neglecting platforms that made it what it is today (mobile especially), they are focusing on the PC platform too much and they lost that race from UE a long time ago.

 

I was thinking about your post today and I it really hit home how devastating nanite could be to unity. Unity has gone very wide trying to grab so many markets, film, car, architecture etc. 90% of why these companies come to game technology is real time rendering. I don’t think unity can offer anything better Nanite/Limen and I think that is a problem

 

Better to focus on making the game and content. The years go by very fast and when you least expect it new tech catch up!

 

I doubt that UE5 uses Mesh shaders for Nanite, since they'd have to discard any GPU older than turing arch (below RTX 20 series), same for AMD Vega, not to mention, I don't think they're supported for DX11 and below either. So, to my understanding so far, Nanite uses compute shaders for the most part.

Also, mesh shaders are basically doing all what VS, TCS, TES and GS are doing in one stage, smth like a compute shader. So LOD and anything that involves Vertex and Geometry are done in one single stage than separate multiple stages which results in memory efficiency.

 

For Nvidia the support starts with Maxwell (GTX800-GTX900) and for AMD it starts with GCN 1.0

 

There is already someone programming a nanite solution for Unity, this is one of their tests, it is an interesting job

Thanks NanoTech

 

Kind of curious though. If you're in the market for nanite...why not use nanite? Radical idea, I know.

 

Roadmap for UE5.1 is great https://portal.productboard.com/epi...roadmap/tabs/80-unreal-engine-5-1-in-progress

 

It is always good that there is competition, at some point Unity3D will catch up to Unreal, but always the type of project will determine which engine is more ideal for that purpose
Regards

 

No, it is not possible for Unity to catch up to Unreal because they are moving in different directions.

 

What do you mean different directions?

 

If your detail is that they're both capable of games then it's not a different direction.