So, Nanite

Because people talking about cool ground-breaking features of both engines, in general. And I agree that the DOTS is the way to compete with UE5. I'm not defending any engine, worked on both (mostly on Unity, but still). Also, you should understand, that being good at everything is pretty challenging. Unity still the leading engine for 2D and mobile, with no doubt. It is also very fast at working/prototyping in general, so better for indies.

UE5 with no doubt at this stage is better in graphics, two major techs (Lumer and Nanites). Unity has multiple real-time GI solutions right now, but I think they don't know on which they want to focus, Enlighten, RTX, and Realtime GI (fake RTX). Also, I failed to find the process of texturing these high poly meshes for Nanites because Substance Painter obviously will fail at 1 million triangles.

Control Rig is basically Animation Rigging, with visual IK bones. But the sound engine looks nice, can't tell if it is useful to have oscillators and basically the whole synthesis inside it, but we will see.

I also wish Unity will continue enhancing the Visual Effects Graph because UE Niagra is the same GPU-based particle engine, but more advanced.



Anyway, I wish success to the Unity team. It is difficult to be the best in every direction, but I think with the release of DOTS the puzzle will be completed and we all see how good it is.

 

You need incredibly optimised draw calls and scene queries / culling, without you can't feed the GPU fast enough. In some cases you will be draw call bottlenecked or even in other areas.

DOTS makes rendering faster, simple as that. If you want one single blob with millions of polys you can probably skip DOTS.

Most games will want thousands of dynamic things, but settle for very few in Unity at the moment.

 

Quixel can handle it, and that is what Epic is pushing, since they bought them up. (Mari is another option that was used a lot in the film industry before substance even existed, but the painting tools aren't anywhere near as helpful.)

 

It looks like mesh shaders, but it's not mesh shaders. They have docs on it up. It's a combination of several things, some of which you can do with mesh shaders:

- Breaking up a mesh into chunks/meshlets.
- Decimating at the meshlet level.
- Meshlet compression.
- Streaming and stitching meshlets at different LOD levels.
- Compute-based micro triangle rasterization.

The last point cannot be done using mesh shaders because it bypasses the GPU geometry pipeline entirely.

 

As far as I understand they use software rasterizers to avoid overdraw with tiny triangles. But in the end they try to maintain a stable pixel/triangle ratio throughout all frames. Like a triangle or a few per pixel, which comes down to roughly 2 millions triangles for full HD which is the resolution they are targeting (not counting upscaling now). Or 3.7 millions for 2k. Aren't these the numbers that could be handled anyway by the raw computing power of 2xxx and 3xxx nvidia series that is adviced to run their demoes? Provided that other steps would be taken care of by mesh shaders.

 

The problem isn't the amount of triangles, it's their size on screen.

GPUs always need to shade at least 2x2 pixels, and simply discard the results which fall outside the triangle. When your triangles are smaller than a quad, the GPU is wasting resources processing invisible pixels.

Check this out:
https://www.g-truc.net/post-0662.html

Mesh shaders serve to increase efficiency in the geometry processing stage, by enabling new culling and data sharing scenarios, and making it easier to hit higher SIMD lane occupancy rates. But it doesn't change the fundamental fixed function triangle assembly costs when triangles are too small. You can at most avoid sending triangles down the pipe which wouldn't be rendered at all, but the ones that do make it to the other side will cost the same as before.

 

Hippo, but DOTS is barely out of infancy...it just is not ready...can we seriously say that the tech exists? I can't. I can't even recommend it. Last I checked was about a month ago though.

 

So, theoretically, tessellation (for keeping triangles always the right size) + effective culling (mesh shaders or something similar) + enough computing power (to draw what's left on screen) should be enough for the output compared to nanite?

 

http://www.elopezr.com/a-macro-view-of-nanite/
Good post

 

Well, I hope Unity has something to show in the upcoming events because I feel they have gone unusually quiet around here

Maybe they could surprise us with something

 

It probably has something to do with the fact that publicly traded company cannot share news willy-nilly or they get into trouble in no time. (It can affect their market standing and that's bad)

 

Dunno, Unity went public like a few months ago but Unity has been like this since the release of Unity5, so...

 

I'm not sure we're talking about the same thing. Unity in general (many employee, including Joachim too) were very active on the forums and on other channels and we constantly got information where they are in development and what to expect, what features they are even thinking about (well, most of the time).
This has changed lately and in the past months, not years ago.

 

I've always felt that the Unity devs in general are not open to discussion especially about what is going on and stuff, but that is my subjective feeling so I can't really say how it is for you mate. Regardless, I'd like to hear from devs if something is going on at least. I can see Nanite and Lumen becoming the standard people expect for Realtime GI and RTGI very far from being a standard due to performance reasons. I just want to know how my future plans need to change and get an honest information about whether Unity has a solid plan for feature parity other than SSGI+Probe Volume which at this point is already outdated before its release.

 

No, ECS is the efficient computing part you are thinking of scripting/coding-wise. Well, at least the mainstream thinking of it. DOTS itself is the efficient computing for the whole engine. It means Data-Oriented Tech Stack. Basically, the whole engine is being rewritten to work on DOTS.

To be honest, if you Compile everything that's being worked on for DOTS and that's being said over the years from Unity, you will see that there's an equivalent to all that UE5 will offer. It's just that it will take some time as the whole editor is being rewritten as opposed to UE5 being additions to UE4.

 

Don't know mate, UE5 seems like a generation ahead even with the prospect of DOTS and all related upgrades that could be possible with DOTS.

Just compare the best looking Unity game (Tarkov? RUST?) to UE4 stuff...and then UE5...

I used to be a huge DOTS fan...drooling over the idea...but even for me there is a limit to how long I can chase that DOTS unicorn. Also, I think DOTS used to be this golden goose that we all sought after, but after all this time, my evaluation is that it is a good backend system structure to accelerate the Editor and maybe, if you need extreme control of the memory, then for actual games. But in general, it doesn't add much to graphics, it just optimizes the CPU workload in certain situations. Nothing more nothing less. If you need more frames, I think turning DLSS on will be much faster, simpler and effective. It is a bit sad, but that is the reality.

 

Maybe, but they will be mistaken. Besides, it's a preview tech at this point, I wouldn't go that far just yet (although I seriously wish they succeed with this, it would be awesome)

They brought back Enlighten, it's more than clear answer in my book. For better or worse, there is the answer.

I think both companies are doing great work. Obviously for historical reasons, Unity is in serious disadvantage, being the significantly younger engine and still having more technical debt and more complicated situation with support for many platforms. I know it's not popular view nowadays, but I don't see Unity failing, on the contrary. They'll get there.

 

It's not about the chase or DOTS being shiny. It's about the new foundation of the engine and the new tools and workflow it's going to bring. Devs are working hard right now to have that solid and bring us the tools, we just gotta wait.

I know seeing Unreal doing all this new and advanced-looking stuff turn heads to Unity wondering, what's up? They are on two different paths though. For Unity, they are "behind" because the foundation isn't there so they are working on that. If You remember UE4 was a rewrite of the Engine but Unity 5 was just rendering upgrade to PBR and basically patchwork after patchwork to get here where we are now. So Unreal got a new core/foundation while Unity didn't/don't have one yet.

What we have right now and to look forward to Until the new Foundation is done, is just current Unity being polished and bug fixed until the New Editor is ready. The most "new" features we will get are from SRP and some other packages for a while. All Major efforts are on DOTS.

 

But this preview tech is already changing the industry. If some dude can achieve realtime GI'ed scene in 1 hour that changes things. And once this happens, and it did, things don't go back to way it was. Devs are not going to sit and wait a whole day to bake maps...it is just not happening anymore.

Think of PBR...remember how many people said that it is an overkill? That similar effects can be achieved with cubemap specular?

Think of volumetric effects and GPU particles.

Think of motion captured animations vs hand made.

Once the tech becomes a viable option, its like opening a pandora's box. The world just moves on. And this is why I am worried and want to openly discuss what we are doing with Unity.

 

And at this point it will definitely be too little too late.

Unity always had a certain simplicity going for it, if they lose that (since DOTS seems to complicate things and Unity seems unable to find an elegant implementation), people will start dropping Unity faster than a certain CEO dropped his Unity stock.

 

Yep, they just need time. I think if they properly laid out things people would have a better understanding of what's happening. The reason I even know the engine is being rewritten and all the tools are being transitioned to DOTS is because I keep a close eye on things almost religiously. Still I miss a lot of things.

If they came out and say

This would help everyone understand where they stand and what to really expect moving forward.

This info is kinda there but broken up into several Blog and Forum Posts, so if you haven't seen most of it things seem like they are at a standstill for Unity.

 

I'm too old to participate in the premature over-hype any preview tech, no matter how awesome it looks at firs glance.
Things usually don't go back how it was, but we aren't at the other place just yet so no need to come back. Or something.

You're mistaken. Plenty of people are using "old tech", despite the fact that the "new tech" is available for them. It is a matter of deliberate choice. It is adorable that many people think that "everyone needs it, just because I do" or "this is such a basic feature because I am dying to have it". Tech doesn't work like that.

And I still think, this tech is awesome if Epic makes it work on a sensible level of production readiness in their engine.

Not with these words, but they basically stated this on their keynote on GDC Showcase. Obviously on more marketing-ready language, but this was the meaning and the message.

 

For some reason, this video got me much more excited than what unity showcases now

 

Ninja, I don't know, but it feels like you are equating this to kids wanting a new toy. If you are then I think you are looking at this very differently.

No one here, at least me, is moaning or whining because we want this new toy. This is not about that. And even if it was, there is nothing wrong about it. I mean I did say that I was a DOTS fanboy but that was just me and DOTS

What this thread, and us as a group of devs are doing, or me, as I don't speak for everyone, is enquiring the company that I do business with to see and determine if my future investment needs to be continued as the competition have up'd their game and my side (Unity) seems strangely quiet.

Now, if you come in saying "plenty of people still use old stuff" or "don't be too hasty", I agree with you and I share the feeling.

But, I cannot help feel that we are here for different reasons. I, personally, am spending time here because I am desperate to determine whether the next few years of my consecutive projects need to be Unity based or Unreal based. If Unity has plans to match features, I will most likely stay and the benefit off of my experience and the existing bank of scripts. This will accelerate my progress by a considerable amount.

On the other hand, if I am spending the next 3 years making another open world game in the current Unity environment, it is most likely that I need to jump ship, discard my experience and bank of scripts as I will never be able to best the previous project. I will probably get cancer if I need to work on another one. The pain...it is real.

So really, I am waiting for the devs to say something, share something...something so that I can make informed decisions. Even an announcement of an announcement would be appreciated.

 

Yeah but this is not what current HDRP is about. This is just blatant marketing that we should ignore

 

If it's blatant marketing then so is Unreal's demo...
You can achieve this quality, it just takes either a lot of work or money to buy high quality assets. Most big budget games that are going for high quality graphics are either going to be using their own long-time studio engine or Unreal because of it's proven ability to handle these big AAA games, so Unity is not going to have many examples. But you really can achieve quite realistic quality with Substance or Quixel assets, just like Unreal is doing in their demo. It's the photogammetry that's really doing most of the work here if we're being honest with ourselves.

 

I didn't think like that, if it came through like that, sorry, it wasn't my intention. On the other hand over-hyping any tech, especially before release is just harmful. "Industry-changing" and all. I'm not saying they won't pull it off, but I still remember the kite-demo and the "revolutionary open world editing capabilities"... ended up being an average subscene solution. So I will hype it when I have it in full. Until then, it is awesome, but I won't base any business decisions on it.

I understand your position about needing information, but you also need to understand that Unity can't discuss things freely anymore for legal reasons. Obviously you can ask away any time, I just wouldn't expect clear answer.
You have a business, you will have to rely on your best judgement. Besides, always use the best tool available for your problem. If you're making open world games and you're comfortable with Unreal, I say you should go for it, simply because I highly doubt that Unity would give you open-world-related toolset. They stated multiple times that they don't really think it's the direction they want to go. Obviously ultimately up to you.

 

Almost everything in that video we currently have in the engine for HDRP except sky occlusion probes. Well, we kinda have this in the form of Probe Volumes but not exactly the same thing since Probe Volumes store GI and not just sky occlusion. They aren't enabled by default since it's still being worked on but you can enable it if you want. Then you can Bake GI without lightmaps and faster than baking full lightmaps. GI will be stored in the probes, it's not dynamic though.

 

Not this time. Unity's demo was about pre rendered, ray traced + fx that is cinematic quality which can't really be used in realtime. Promoting that as HDRP is definitely marketing. Playable frame rate is just not achieveable.

Unreal's Lumen and Nanite demo is not. It is meant for games, realtime and there are already hundreds of people making vids on Youtube with various hardware spec. So, no, while Epic may have tried to oversell their new tech, it is not blatant marketing. Probably because they are really confident about this one.

Yeah, but you can't really make games with it. The performance requirement goes through the roof with those settings. The tessellated trees, terrain, Raytraced GI? I get 10fps in my 5950X + RTX2080 @ 1440p. How do I know? Well I tested a few weeks back.

Hence the prerendered HDRP promotion material. Or a very small scene with one super high detail. Either hides the fact that Unity can't handle realtime GI and high poly mesh. Simple. Unreal on the otherhand, Nanite and Lumen.

There is nothing really to debate whether Unity can compete in this regard. There just is not. We are really discussing what Unity will do in the future...

 

unfortunately from my testing HDRP has serious CPU overhead, I feel like it'll be tough to make big games with it. It's supposed to scale well with lights etc, but the CPU seems to be a serious problem. I still love HDRP, it gives you the opportunity to use unity for what you love it for
(for me: doc, API, c#, editor customizability) and have many more engine features and high-end graphics.

Unreal can handle much bigger worlds, much bigger landscapes and more details In generaland this is UE4. Ignoring any improvements UE5 will have on release. Unity HDRP is more fit for small-scale games, and if pushing it, maybe a bit bigger. That's about it.

EDIT: HDRP made big improvements to CPU performance in 2021.2 -- honestly I've been using it for a while now and i'm quite surprised and happily so. Much better than previous versions.
Also GPU performance was improved.

It's state is much better than previously stated. Still quite a bit to go to scale better but it's looking good.
Here for 2022, looking like there will be big performance improvements in 2022.2 -- hopefully it actually lands

 

Those are all default HDRP not with raytracing. So it is possible.

Nothing about raytracing was mentioned. Even if you did use raytracing you can still have decent performance if you have the hardware and set things up right. Even I've used Raytracing with my GTX 1080 and got 30 fps at 1080p. It's not even built for Realtime Raytracing.

Example - https://forum.unity.com/threads/unity-experimental-hdrp-dxr.656092/page-26#post-7189825

Raytracing is a few years off for full performance and quality in almost every engine. Metro managed to pull it off with the Enhanced Edition but they probably had to do a few rendering tricks, to get it all to work. Actual Realtime Pathtracing is the end goal for Realtime Raytracing. It's a few years off in terms of hardware and adoption, as well as software. Right now everything is being tested.

Well, it looks like roses and rainbows from outside. You should test it yourself. I have and with a GTX 1080 GPU and a Ryzen 3600 CPU. On a default new Third Person Template project in UE5 with the new features enabled. Mainly Lumen, Virtual Shadows, and Super Resolution all make my PC crawl to a mere 35 - 40 fps at around 1080p in the editor. Just the default scene. Nanite on the other hand doesn't hit performance. If I disable all of that I go back to something similar to UE4 at 160 or so fps in UE5, while the same scene is 200+ in UE4. By default, UE5 is more taxing and the new features are really heavy as well. The Only people getting 60 FPS are the ones using the new high-end cards. Both Nanite and Lumen have their limitations. Like I said before and I know this is a Unity forum but test UE5 yourself.

So if you are talking about performance then I'd say that's way off.

With the demo that was presented and available, you can see the high requirements for that. At 1440p it's barely hitting 60 fps on the top of the line cards. I'm talking about RTX 3090 and RX 6900XT so yea. The only tech that really helps with performance is Nanite and this is basically a software version of Mesh Shaders so it doesn't rely on DX12 or any hardeware. Everything else is a workflow upgrade and performance drain. The Grass is always greener on the other side.

 

Grim, I think your idea of Raytracing is a bit missing the mark.

For one, experiencing and leveraging RT in pre 2000 card (hence your 1080) is near impossible. If you were getting 30FPS 1080p Raytraced GI environment, then either scene was super simple or there is magic that I am not aware of. 1080p raytracing gi should pretty much get you something between 5~10FPS given that the scene is complex enough like an actual game - not a test scene.

For example, I struggle to maintain above 45 FPS in my 2080 with Raytraced GI in my bench scene (about 6~9M poly) so I am almost certain of this. The Buildmark video that you posted, while I am not sure the hardware spec, dips to 20FPS in 1080p when there is a building and even then the environment seems way simpler than a project that would use Raytraced GI. What I am trying to say is that Unity Raytraced GI is near unusable at the moment due to performance reasons.

Now, as for Unreal 5, which i have been working with for the last 4 days, the performance is outstanding. You said 1440p Raytraced GI barely hits 60 and you are complaining. But look at that scene. It is massive and the detail (not just poly wise but in every aspect) is definitely above AA open world map. I mean...do you know how impressive that is? I can guarantee you that there hasn't been a game released that uses Realtime GI in the scale, meaning this is a playable demo that is way beyond what open worlds do in terms of detail. A stress test level of sort. That scene, even if it was without Nanite and was with Unity-level low poly, Unity would struggle to hit even 10FPS on 1440p. That demo is as fast as Realtime GI can get and you are not impressed? I am scratching my head here mate.

In a general setting, where the scene is not meant to stress test the GPU out like crazy, there are many videos where people easily hit above 100FPS in the editor...in the bloody editor. Here is one video.

If you are still not convinced how good this is, then turn on Raytraced GI in HDRP test scene in the editor and see what FPS you get. Actually I will test it now and here are the results

HDRP11, Unity 2021.1.9f, AMD5950x RTX2080, 64GB RAM, NVME SSD

HDRP Sample scene in editor, baked lighting disabled, everything else untouched, DX12, 1440p
135FPS by the golden ball
122FPS by the plant

HDRP Sample scene in editor, Raytraced GI, everything else untouched, DX12, 1440p
53FPS by the golden ball
45FPS by the plant

The sample scene has about 0.5 mil triangles which means it is not even close to a normal project.

I will port the Unity HDRP sample scene to UE5 this weekend (hope it wont take more than 30 minutes) but from the looks of it I am confident that I can get around 120~130FPS in UE editor.

I am surprised that there are people who think Lumen is not impressive....especially given the Unity Raytracing GI performance...

As for Nanite, people are getting 80FPS in editor with 500 Million triangles...500 Million...That means age of LODs and detail maps is over - like completely.

 

As I've mentioned that's on higher-end GPUs not really what a lot of people are using. (The general consumer playing the games)

I mean, the video you are using as an example is showing how hard Lumen hit's performance. There's also no info on his hardware so there's no real measurement to take from his video except it runs from 30 - 60 fps most of the time in a simple scene and 100 fps (When he's looking at a wall). I'm just relaying the reality of things having used it myself. Lumen didn't claim it was faster or fastest just that it works and works better the more powerful your GPU is.

It's not that people aren't impressed, we've all seen realtime GI in a lot of different implementations, so it's nothing new just another implementation that's more aligned with the current and future generation of hardware. Nanite is the more exciting tech as it doesn't rely on DX12 or the new hardware to work. Software mesh shading similar to what Infinity Ward Implemented in Modern Warfare 2019 but on a larger scale. Nanite in UE5 is just getting in line for the next-gen GPU pipeline but on their own terms.

This is where things come back to DOTS and Unity since if Unity were to do something similar it would be with DOTS and the Hybrid Renderer or whatever it will be by then. Or just fully embrace mesh shaders as it will be used for Vulkan and DX12. Although I'd expect Unity to have their own general solution (like Nanite) just so it would cover a wider range of platforms. Like there's Shader Lab for Shaders

 

Obiously, we disagree on many things like:

Realtime GI being seen in many different implementations -> where else that we find this? Unreal and Unity make up almost 90% of engines that we can buy and use so I am not sure why Modern Warfare's implementation even matters. Unless you have the MW engine license? Unity as I said can't do it, and UE just made it. I don't really see how this is just another good implementation of Realtime GI...this is pretty much the only Realtime GI that is viable and it arrived like 5 days ago.

Lumen's performance and that it is only for high-end GPU -> Lumen is the only champion that features both usable in performance and quality. If this hit is too much then pretty much every other implementation is gonna be harder. Also, I kind of find 2080 to be mid end, slightly above average. It is just PS5/Xbox Series X level. Not high end, but more like mid range. I don't think anyone playing open world, realtime GI games are trying with a weaker GPU - it won't work anyway. Look into Cyberpunk, or Days Gone and etc AAA games and specs. This is subjective, but when we are talking Realtime GI, we are not talking low/mid spec really.

I'm just relaying the reality of things having used it myself -> You and I seem to have a different idea of what is considered "high-end", and seems like my standard is geared towards a bit higher end environment, and yours a bit lower, and assuming the GPU of most people, your stance is probably not a bad one. But, yeah, I wouldn't say Lumen didn't claim that it is faster, because it is faster, by a huge margin. If you don't agree...then try to replicate Lumen by turning on Realtime GI (SSGI or RTGI) on Unity and see. You will quickly realize that it is not just "it works" but also definitely "faster".

 

Does Unity(right now) work on these smoothing lod techniques? I also think they(unreal engine devs) will introduce doubles to support creating of big worlds(not currently available but i read problems are with their networking but they plan to add support) so is Unity working on this too? It would be nice to hear word from Unity devs ...

 

From what I heard by watching Unreal streams they started thinking about possibility of Nanite 10 years ago and by 2017 they had a working prototype, and it's 3 times faster than hardware rasterizer. Personally I don't think that Unity is going to get a magical solution in a year or two, but maybe I'm wrong.

 

They were working on some HLOD technique https://blog.unity.com/technology/u...nting-with-automatic-performance-improvements (which isn't nanite, but still, it would be very welcome) but it was abandoned (?) a year ago or so.

 

Actually it won't be as hard as you might think. Once the tech exists, the risk of spending 10 years on R&D, and realizing that it was a waste of time diminishes, the need to convince investors and justifying the feature is much easier if the tech exists. Especially if your competitor has it and you don't So no, it won't be the same for Unity. Unity won't have to go through the painful 10 years that Epic had to go through. It is just playing catch up. Nothing spectacularly hard - virtualized polygon streaming is not like solving Riemann's Hypothesis. It might require some new teams, resources and careful direction but then again, this should not be like moving heaven and earth.

The real issue is Unity being Unity.

 

I think Unity's a fantastic element in the industry that deserves the chance to manage themselves a little stricter.

In this case it's probably best that DOTS is that fresh start and a totally clean break from mono land and all the legacy before it.

I don't see any other way other than Unity biding their time and not adding more work on their plate. So they could do nanite but they wouldn't have resources to satisfy people using it like an entire supporting toolset.

But if they just treat DOTS as Unity 2 - a separate entity, they can make some hard future decisions about tech without constantly upsetting people with changes and so on. It's a good opportunity and I hope they don't muck it up by trying to please everyone.

Let's learn from history shall we?

 

Yeah they can make Unity in few branches if it is hard for them create monolith game engine - for example 1. dots + unity version of nanite(when ever it come in future and after they fix current problems in Unity) - focus on performance (AAA games etc) and especially big worlds which are future of game programming etc 2. Unity for 2D and 3D small and medium indie projects and maybe there should be mobile Unity or as a separate entity and ofc - focus on simplicity of coding for small to medium indie game 3. Project tiny which also aim(i think) mobile platforms(idk much about PT to be honest). Now i understand if this would be hard for them to split on so many fronts - but also making monolith game engine as dots example shows clearly is extremely hard to achieve...

 

They don't make a monolithic game engine anymore, this was the purpose of the package manager and at this point most of Unity has been compartmentalized.

 

Well from what i understand(correct me if i'am wrong) dots is not in Unity 2021 because dots needs core-Unity changes that are bad or incompatible with classic Unity(MB Unity) - if package manager can handle everything then we would have dots in 2021 - but this is pure speculation since devs did not confirm the reason why dots is not in 2021 so actually you might be after all right

 

The package experiment has more or less failed. The "verified" packages, which was a positive step, is more or less them admitting defeat.

 

I'm not sure how this makes sense, the package "experiment" has not failed at all and has been a net positive, allowing you to have a much leaner Unity Editor experience by being able to exclude most components you aren't using.
"Verified" just lets you know it's out of a beta stage... The fact it has its own separate beta phase from the Unity engine's beta phase is exactly the point, it's compartmentalized into its own development phases, thus my initial response to the person claiming Unity needs to be able to work on features separately; it's literally what they are doing now.

 

Totally agree with this.

 

Yeh dots is basically instancing running on multiple cores, aka multi-threading. BUT you CAN NOT just drop a zbrush model or raw scan data in and duplicate it a million times.
Thats just silly to try to compare it to DOTS. Nanite is here to stay and Unity has to match it to compete on next gen for sure.

 

I get what you mean but it's not all or none, you'll need great CPU performance for great GPU performance. Nanite is one way of achieving vast quantities of 3D geometry, but insane CPU performance is another way of doing it.

You need DOTS or similar tech to cull and manage all these instances. You won't be able to in pure C# land, it's simply nowhere near fast enough. A fast renderer, or custom renderer is often limited by it's supporting CPU performance for example HDRP could be faster, but isn't due to being chained to a legacy engine.

DOTS is not just "hey we can hammer the CPU and max it out efficiently" it's also "hey we can feed the GPU this fast with this much custom data".

 

Thx for this info - also here is their github page : https://github.com/Unity-Technologies/AutoLOD
Unity - revive this project

 

I disagree.

 

The limiting factor for many of the techniques employed by nanite isn't CPU execution speed, but the overall rendering architecture (SRP and the whole lot of stuff that's still done in the C++ sandbox), which is very behind the times. They are still catching up with instanced rendering with DOTS hybrid renderer, while others are moving to meshlets and GPU-based pipelines.

Nanite, for example, requires DX12/Vulkan (and Metal, eventually) because it's built around things like low level GPU memory management and indirect execution. Meanwhile, Unity's implementation of these low level APIs is still quite conservative (which is why DX12 in Unity runs slower than DX11).