Discuss…
1 Like
You’re late to the party. We discussed Unreal Engine 5 and its implications when it was announced back in May.
4 Likes
If we do get it it will take years and years.
Though I would guess nanite is too demanding for VR@90 FPS anyway for current gen hardware
I was in the boat of nanite will probably not be as amazing as they say - but I have been using unreal a ton lately and I am just blown away with the features and how high quality they are. Hair for example was just easy to do and get high quality results. Same with dynamic destruction. Ofcourse I am only using it for evaluation purposes currently, so I have yet to try and make a large project with it and so take what I am saying with a heavy dash of salt.
But I am really optimistic now about nanite based on my usage of the engine so far. If it comes out and you want that feature, just use unreal instead - no need to remain on a single engine a lot of people are using multiple these days due to how uncertain the future is around who will end up being king
6 Likes
Yeah, it’s evident that there is a big plus that the engine makers also make real games with their own engine.
There is just so much workflow issues with untiy.
4 Likes
I mean to be fair workflow wise unreal is as much as a hot mess as unity, but the difference is that there is a tool for almost everything you need, rather than having to build a lot of the systems yourself.
Also you basically HAVE to use blueprints for certain things (I mean you can do it all via c++ but certain sytems have been built to be basically obfuscated to c++ and only exposed to BP so it requires a ton of source code inspection + editing to get at certain bits as they will have no docs as they intend you to use the visual scripting) which as a programmer does not sit well with me.
BUT overall despite still learning the engine I am working faster in unreal than unity, which is crazy to think because I always said to people that unity was the engine for making things quickly and its turning out that is not really the case at all despite what I had thought from previous times trying the engine (it seems to have massively improved since last year in that regards).
I am very impressed with it overall, I can see myself using Unreal for games and Unity for non-games projects I do through my company - but its likely over time as I get to grips with Unreal I will just swap to that entirely I think (at least until Unity gets back to being where we need it to be in terms of stability and workflow and unity themselves working out what their general plan is moving forward internally).
Also until the nonsense with SRPs ends and they become as usable as in-built and the strange decisions to hide parts of it gets reversed, its easier to work with unreals shading pipeline currently even though I know way less about it compared. Right now with unity SRP it feels a little bit like every new version (8,9,10 etc) its basically a completely seperate thing with massive changes even still happening. I also have yet to see asset store developers and tech artists get to a point where SRPs are not fighting against them…
6 Likes
I havent used Unreal but I would bet the world creation tools arent as bad as the Unity ones. And I talk mostly about lightmapping here. Lightmapping is a fuckign hell in Unity.
This makes me curious… As I don’t think Epic changed anything significant in general engine workflow last year. Sure, there are more and more systems and editors ready out of the box. LiveCoding is fantastic, but they added it earlier.
I would assume that just your next approach to the engine is different? Or learning materials you found are better?
Or maybe something actually changed for newcomers and I don’t see it because I use Unreal since UE3? 
I’m really curious about what has changed from others’ perspective!
My perspective is that UE4 (not UE3!) was always easier to make 3D games than Unity. Not at the beginning, if you’re new to gamedev and try your first 3D game - Unity is much easier to create your “3D hello world”, but… to me, it was always like “easy to learn basic things, insane to use it every day in the long-term project”.
I have no idea why these threads get off topic so easily. This is supposed to be about Nanite, not another Unity complaint thread.
The real question is whether or not you actually want Nanite. How does Epic plan to allow LOD-less meshes without those LOD-less meshes taking up their normal disk size? Meaning, your one level prototype is going to be gigabytes large. Do you want that? Do you care?
And in addition, this simply places more burden on artists to produce high-res art…and that’s arguably the worst part of AAA (in terms of ballooning budget and customer expectations).
None of this sounds like a good thing.
1 Like
Epic shared some stuff about nanite and it looks like early technology. It does support only opaque objects, and complex meshes (like trees) are causing performance issues. Memory is a limitation too, artists have to be careful with how rich worlds they want to make.
The biggest unknown about nanite is performance. How it will perform on older gpus and weaker cpus? Can I use this for cross-gen title or is this tech purely for high end pcs and nextgens?
In my opinion, it’s still the case. Unity does have better documentation and a lot more tutorials about basics. The UE is the opposite, if you have 0 understanding of basic concepts in game dev or/ and programming, you will have a much harder time. Once you will learn the basics, the workflow is much better in every possible way with ue4.
If you are making a more generic 3D game, then the ue4/5 is the way to go. The main strength of Unity is c#. It’s easier to make more unique, innovative systems and mechanics.
The problem is never (almost never) your own systems. Problem comes in rendering perfomance. And the tooling is very bad for optmization these things in Unity. Heck., alot of the things we need to optimize should be part of the automatic process. Like spatial tagging of lightmap atlas etc.
2 Likes
I don’t know much about the technology but I do know artist becoming focused on high res zbrush sculpts, prioritizing “artiness” over efficiency is just about the opposite way I would want to spend my money if I was running a studio. Massive time sink.
2 Likes
This is all true, the nanite seems to be directed more towards Disney and their movies than towards game developers due to its unfinished state.
And about the size, I think there is a way to overcome this issue. Instead of using rocks with millions of triangles, developers could try to use modifiers to simplify the original mesh. If the math for simplifying the mesh is proper, then there would be no visual problems, the engine would generate lower LODs, the size of the game is lower. Hopefully at some point epic will try to do it themselves. For now, the nanite is a pure gamble with all these issues.
1 Like
You misunderstood my OP. Let me re-phrase my OP -
Now, 8 months later, Unity has had time to react to UE5’s potential, has Unity mentioned any plans to compete with UE5’s Nanite system?
Makes sense now?
No, the answer is no, Unity has always been playing catch up and with the current rate of development, they always will be.
Please buy Pixyz now for an additional price, so you can have automatic LOD in Unity, a feature Unreal has had for years for free.
1 Like
No. Why would they? See my previous post.
Nanite has years of research behind it and Epic won’t publish the implementation details until it’s released, there’s no way Unity can come up with something on the spot.
Rendering-wise, Unity has been playing catch up as @AcidArrow said. HDRP ray tracing, for example, is still far from being usable for game deployment (since DX12 support in Unity is still experimental and very unstable) and is not even available on PS5 and Xbox Series yet.
Heck, they are even playing catch up with themselves with URP (which is still missing key features from built-in, two years in).
Unlike Unity, Unreal is actually used by AAA so it needs to have cutting edge features to stay relevant in that space. Keep in mind that Nanite is targeted at PS5/XBS games, to take advantage of the new generation hardware.
Also, Nanite isn’t “LOD-less”. What happens is that the LODs are virtualized and are streamed at the patch level, enabling the usage of vastly higher quality models compared to the traditional “each LOD is a full separate mesh” technique, combined with compute-based micro-triangle rendering. Essentially, it’s like mip-mapping, but for geometry.
It doesn’t work for “porous” geometry like foliage because it “merges” nearby triangles into larger triangles to create the “mipmaps”, which creates visual artifaces for non-contiguous geometry. This same problem happens with all LOD generation tools, BTW.
4 Likes
I’m not an industry expert or anything, but I find this hard to accept. High-res art is already created, even in lower poly models. Characters are still sculpted even though the end asset is much lower poly. High-res art is used to create normal maps, textures, etc. (this is my understanding)
It seems to me the burden for developers would be the process of reducing the high poly art and optimizing assets - which is the aim of Nanite to resolve.
Though I do agree with the space and memory issues that Nanite inevitably has to deal with.
1 Like
All of this makes sense.
Storage usage depends only on the maximum quality your models will have when viewed up close. You can use Nanite for not-so-absurdly-high-quality models, which won’t look as good close up, but would still take advantage of seamless geometry reduction and streaming when far away.
3 Likes