Graphics Nano Tech - Similar to Nanite - High detailed rendering for HDRP / URP and Built-In RP

Hello!
I'm working on a Unity tool called "Nano Tech", which is similar to Nanite.
If you like to support it, you can follow the crowdfunding campaign:

https://www.indiegogo.com/projects/the-unity-improver-nano-tech

Thank you!

 

This is neat and looks great!

A few things you should do in your campaign:

• explain how it differs from Nanite ie why choose yours over theirs (or Unreal for that matter). And perhaps why you started this when Nanite came out - coincidence? Not satisfied with Nanite? Just curious to try it yourself and went further than you expected?
• benefits - if I were to support a campaign like this I'd want to know what I get out of this. Right now there isn't much to go on, so why would I support it? Do I get free access to it? Etc etc.
• timeline - WHEN it will be ready (alpha, beta, release) is also important ... otherwise I'm thinking you're just going to work on it on or off without clear goals or maybe not even a plan.
• price point - especially when comparing to commercial solution like Nanite, I suppose your price point will be lower. But you're not mentioning pricing at all.
• Watch the video: I found the button more by accident than anything. If there's a video I'd expect it to be embedded at the top of the page and playable with a single click.
  • skip the text at the beginning, get right into the visual part - I don't want to be told for 10s that it's going to be awesome! SHOW ME THE AWESOME! You can easily lose a significant portion of viewers with such an intro or they'll skip ahead and miss relevant parts.

 

I can't speak for Christian (the creator of the tech), but he's been uploading videos to Youtube about the tech for the past 6 months. Reviewing his LinkedIn -- it looks like he's been a professional graphics programmer for over 10 years, and most of that work has been in Unity.

As for why he's making his own Nanite-like (and Lumen-like) tools for Unity, my guess would be that Nanite Virtualized Geometry is for Unreal v5 and many of us on this forum would prefer to have the same graphics options available to us over in Unity so that we can continue to develop in C# and stick to the rest of the Unity toolchain.

Based on everything I see from my quick review of his work, he's got my support.

 

H! This looks nice, congratulations!
I have one question so far: what do you plan to do with shadows? As Nanite developers pointed out in one of their talks, it doesn't make a lot of sense to have such dense geometry unless you also have per-pixel ultra high-res shadowmaps that could reveal those tiny geometrical details. And indeed, the screenshots and the video you posted kinda show this issue, I think. Neither the shadowmaps nor the AO debug view show any shadowing from micro-details. I'm especially interested in this because I haven't found anything related to the direct shadows in your plans.

Just to illustrate what I'm talking about here:

Spoiler: UE Example

Nevertheless, the tech looks really cool! Very excited to see this

 

Couldn't agree more!
Though for the uninitiated (like myself) such background infos should be in the campaign page.

Also, I was mistaking Nanite for Octane. Nanite rang a bell and the first thing I had in mind was Octane. I knew Unreal had a tech like this but couldn't remember what it was called.

 

Nice to see that at least some single private people seem to be more diligent than Unity-guys.

 

Thank you for your feedback guys! I really appreciate it!

@SteffenItterheim: I will try to add more details later, the pre-launch page of indiegogo is a bit limited, also it is supposed to just give an overview of what is coming, of course there will be much more infos on the main crowdfunding page, after launch.
@Passeridae: I already did some tests with contact shadows (which are based on the geometry depth pass) and the details are visible, but I also like to try deep virtual shadow maps like in UE5, I already created a virtual texturing system for NanoTech so it's basically possible to use it to create deep shadows, however it makes sense to wait until Nano Tech is in a beta state.
@BetaMark: Thanks for the nice words, you get your money later
@Cascho01: I think, the Unity team is just too busy with all the other stuff they do. The main reason why I still stick to Unity, it does not crash as often as Unreal does, just my personal experience. There are other points too, but I don't want to start a discussion here.

 

Well, what else may be more urgent than a Nanite-pendant? A Lumen-pendant, ok.
However....

I´m a bit curios if a single private person will be able to establish a working Nanite-pendant for Unity.
I wish your solution was so good, that Unity-guys will give you some money (much money) and implement it themselves.

 

Well, I already did a proof of concept that it works, but it's currently not optimized, however, I think I can manage this too!
The crowdfunding budget will be used to hire some freelancers to help me to make the development faster and the tool even better. So it will not be a one person project anymore, but I'm ok with that

 

Love it! Count me in for testing and possible promo

 

That's incredible!
However I guess the main difference with nanite is:
- no tile based deferred material rendering
- no HiZ Occlusion system
Am I right?

Which aren't a problem since that offer more flexibility to mix and match than nanite (which has a overhead cost with all these systems), but then the question is, will it be ready to be extended with such system?

 

Hi,

Looks very nice, i wonder what is the main perk of this system, versus just have a mesh with a single texture with all the shown textures blended or a simple shader that gets a texture array and blend the textures in some way.

Is it adding extra detail to already defined meshes for example or can handle endless worlds with endless textures ?

Thanks in advance for any clarification on the above.

 

I'd suggest you to read the Nanite white paper, which pretty much explains what the system does. Basically it allows you to use f.ex high poly 3d scans directly in the game, as the system scales down the meshes automatically by splitting them into smaller chunks with a clever automatic LOD system. It also does deferred material rendering, removing the need for gazillions of drawcalls, allowing you to (theorically) use unlimited polygon / material counts.

https://advances.realtimerendering.com/s2021/Karis_Nanite_SIGGRAPH_Advances_2021_final.pdf

br,
Sami

 

Oh i see, sounds great , thanks for the information.

 

My thoughts about nanite on mobile devices:

 

Have you seen the new phones though? They're stronger at this stuff than some people's computers. Absurd levels of performance. And price to match but that's par the course.

To be fair classic rendering + LOD is faster but this nanite stuff is about higher quality rendering and easier authoring than the absolute maximum performance.

 

Mobile is not high priority at the moment, but we will get there, and the modern devices are capable of using compute shaders, which Nano Tech leverages. At some point there will be a generation where mobile phones have the power of a PS5 and people will just connect their phones to their TV, which is partly happening already. I think streaming comes also into play, but most people also want to play games even if they have no internet connection, having a strong independent device will be still important in the future... At least this is what I'm guessing.

 

iphone 12 and 13 alredy support mesh shaders
I think all arm v9 android phones support it too but we will know it only when latest vulkan standart for it will land in public.
So latest phones alredy can do everything ps5 can but in lesser scale like 10 tflops vs 2.5 tflops

Assuming that we dont want to render 4k on mobile only 1k fullhd means that we have something like same perfomance per screen pixel on latest mobile.

So the only benefit of ps5 is raytracing and ability to render on big 4k screen

 

Where can we get to play with it now?

 

I'll second this.

 

I dont think is this so simple, usually phones have severe bottlenecks that make this 2.5tflops drop dramatically in realistic scenarios.

I would not consider any phone anywhere close in performance to a PS5, 4-5 years down the line maybe can match it in lower resolutions is my guess.

The feature set is rather irrelevant, if the hardware is not fast enough to run those features in a meaningful way.

For example my Samsung S8 is around half speed of Samsung S22, but my 5 years old Dell XPS with a 1050GTX is vastly faster than Samsung S8 in demanding effects, like 5-6 times over and not just by a factor of 2 and considering that i use lower resolution as well.

There is just no direct comparison of any kind between consoles-desktop and mobiles, always will take 5-8 years to do any kind of near catch up imo due to the simple reason of compact form.

 

@Nexusmaster Can you tell me how you store the mesh and virtual textures in a project for streaming? As far as I understand it in StreamingAssets? Thank you!

 

Yes you can use StreamingAssets you can also use a Resources folder or you can load them externally meaning you add a folder to your Unity root project and read from that manually.

 

My campaign is live now, please have a look:

https://igg.me/at/nano-tech

https://www.indiegogo.com/campaigns/the-unity-improver-nano-tech

Thank you for your support!

 

Looking awesome so far!
Don't forget DX12 and Raytracing/Pathtracing compatibility!

Unity should just give you a 6 or more figure cheque for this, and then also have you on retainer to further develop it. I like your mention of a Lumen like tool that's on your to do list! That is great!

 

Pennies thrown in your general direction faster than you can say "nano tech"

br,
Sami

 

I'm really having high hopes for all the tools. Best wishes, with the campaign. Tbh, it would be nice if unity the company takes notices and "hire that man"! At the least buy out the tech, assist or something. I'm sure they and everyone could benefit.

 

Hi, we would like a few clarifications before backing up the project

One and more basic, is there an actual demo we can run in a very big detailed world that can turn on and off nanite to see the difference in performance with an fps counter ? I read the tech specs and seems this system is for very specific uses, e.g small maps with high fidelity meshes

So would be great to see how it fares in an actual game, big open worlds, with different materials, transparencies, variable lighting etc in a demo we can tangibly evaluate.

Second what is the benefit if for example the camera see a map from above and see every mesh, would that system still be faster than just batch everything ?

The tech looks impressive, but so far we have not seen anything much of it yet, the rocks shown in the video is a small place that is already very fast to render, the forest scene in your video also shows lag, even for something that simple that can render much faster in any medium range PC.

So is there any plans for a demo where can evaluate the usefulness of the system first hand ?

Thank you

 

Yes, there will be a demo. But there are still some performance issues that need to be fixed before I will deliver a first demo. If it would be already finished, I wouldn't start with a crowdfunding campaign.
The first early access will be probably in 2 month. I will upload a video soon that will show the current performance compared with and without NanoTech. I don't have a big open world game here, making all these demo scenes and videos take extra time, shouldn't I focus on the development?

Without NanoTech:
- Normally you don't have one draw call for all meshes on your map.
- Also your map doesn't have an automatic LOD system, where each object has automatic sub clusters.
(This means you basically can't just render high poly meshes like NanoTech does, if you create manual LODs you still have to simplify the whole mesh, not just parts (clusters) of it.)
- Your map would render triangles that are too small (smaller than a pixel) or not visible e.g. outside the camera frustum. (higher overdraw)
- You can't use VT or you use Unity's VT but then you have to deal with it's limitations.

So yes, I think with NanoTech it would run faster and preserve more details, if you are not happy with it you can get a refund in the first 14 days after you get access to the github repo!

If you have a look at my old videos then you see I already made good process with the tech, the LOD system is getting less "visible" with each new version and also the performance increased a lot compared to the old version, but there is still much to improve!

Old version:

Newer version:

Thank you for your support!

 

You can probably quickly simulate the technical (geometric) requirement of an open world, by spawning random asset in a plausible way and plop a dumb character controller to make a tech demo. That would impress people looking at it technically for low effort. It seems that would further reinforce the usefulness of Nanotech if this is that simple to make a geometrically dense world.

Because even (offline) procedural world still ask for a lot of management that are tedious to optimize, if that work online too (as in real time spawning of level with pre made geometry, not internet online), then that's definitely a key seller. I think that's worth the marketing fall out, can be over sold as almost a one button solution.

 

Hi, thanks for the info

Why ask a demo is because i read in larger maps with a realistic scenario where there is lot of grass and trees, nanite can give very little benefit and has increased overdraw.

So in some scenarios seems to negate the benefit. Thus would be great to have a demo with what is currently implemented to see how actually fares in a forest and grass fields scenario.

The other issue is that this tech is tricky in general, because how renders things, and is a matter of research still in Unreal as well.

This article describes some of the issues
https://80.lv/articles/discussing-the-possibilities-and-drawbacks-of-unreal-engine-5-s-nanite/

Also reading about grass problems, like cant use vertex deformations with it, so again would be great to see a demo with actual application in a game to see if the system is feasible for use.
https://www.reddit.com/r/unrealengi...d_someone_try_nanite_with_fully_modeled_gras/

 

I think unity officials should support him.
good.

 

backed!

also comment,
if crowdfunding doesn't become a big hit, i'd suggest selling the plugin in asset store as clearly marked alpha/beta, during development.. should bring enough sales to fund it.

 

Or just straight up talk to UT and tell them you need support to make something that will possibly greatly benefit their engine. They bought Bolt to make it their own VS so why not something this grand & useful.

 

Hello!

I was wondering how open is the file format and engine? I was wondering if we could baked a point cloud of clusters mesh (at their adequate LOD post occlusion) from a pre determined POV or position? That would help on weaker hardware to just stream that point cloud instead of resolving real time for predetermine path.

 

This sounds like the tech from Media Molecule who created "Dreams"! It's a good idea but also has it's limits.

 

I created a WIP demo build, you can find the link on my indiegogo page:
https://www.indiegogo.com/projects/the-unity-improver-nano-tech/

 

Not exactly, I don't ask for splat as primitive, when I say point cloud, I mean the position of the cluster to reconstruct the scene from the baked pov (with the cluster). It would be closer to umbra but at the cluster level (or that per pixel asset, "perfect culling" someone is selling on unity store).

It's actually a old techniques done either at the primitive level (triangles in very old game like crash bandicoot) or the objects level (umbra), but with nanite inspired tech we have a new "primitive" with the cluster. Ie the "point cloud" is simply the display list of transform for visible cluster after occlusion.

 

This looks absolutely fantastic!
I see that the platinum tier contains early access but it's nowhere stated what the diamond tier contains. Can you give us some information what we would get as a diamond supporter?

 

The demo doesn't run on my pc. It doesn't go past the Unity logo screen. No errors or crashes, just almost instantly goes to the "not responding" mode. RTX 2070 + Ryzen 2700X. Tried several times. Weird, that's the first Unity demo that doesn't work for me. I'll try to redownload, maybe update drivers if possible.

 

nice demo!

nano tech OFF


nano tech ON


for this view

 

I'm really impressed with this too, here's my stats for on/off on really old hardware (i7-3940xm, GTX 1060m)

 

Did you wait long enough? It might take a few minutes... I had to remove the geometry streaming for this demo, because it's still too laggy. So all the data is loaded in this demo, the final version will support the streaming again and load only whats visible on the screen (or needed).

 

Diamond supporter would be like a business partnership, you get more and better support e.g. if you are using a custom asset from the store and want that to work with NanoTech, I will prioritize this implementation.

 

Ok, I think I got it now, you suggest to bake the current cluster positions so it runs faster on older hardware? Of course this would work but only make sense if the Camera is not moving, right?

 

Did you get this result with putting the performance slider handle (on the bottom right) to the left side?
I think the final version will get you 60 FPS if my current estimations are correct.

 

No, this was with it all the way in quality. I wanted to see how it handled on old hardware

 

FYI to anyone here backing this campaign -- when you submit your payment, check your bank statement an hour later. I tried to back the campaign and just noticed now that my bank (a very large US bank) blocked the charge randomly. I'll have to go back into the campaign and back it again after all the attempts and rollbacks clear off my bank.

 

Yes... kinda.

If you allow to access the data to bake, we can implement our own solution like umbra.

Basically for each defined positions in any spatial volume hash, we would get the cloud points, then stream that whenever the camera enter the volume.

 

I'm really sorry with these indiegogo issues, I'm also not so happy with it, that is why I added the github sponsorware, which is basically the same, but you can pay with PayPal. (Sometimes you have to refresh the PayPal button, but then it works.)