Unreal Engine 5 Feature Highlights | Next-Gen Real-Time Demo Running on PlayStation 5

I think some games will keep optimizing even is there is a feature like Nanite, more optimization would mean even richer environment than what the demo was showing.

Of course, lower specs will always have to be targeted on PC and mobile. That wonā€™t change. When the 30xx cards come along and ray-tracing becomes more optimised, things will advance again. There is no ray-tracing in this demo(not much need as itā€™s mostly rocks), but the end of this year - post 30xx release - is going to be an exciting time.

Iā€™m lot more impressed by this million polygon models demo than most Nvidia ray tracing reflections.

I think this new features are converging with one Epic goal, that is to provide game creators the ability to produce games reaching some triple A quality. Like Quixel assets, gaming or film high rez assets usable directly in the engine.

Apparently Unreal 5 will be able to scale down to mobile as it will provide the tools to scale down million polygon models for lower hardware. Iā€™m not sure that will run faster than games using normal maps for example, i guess the tools will automatically UV map and bake normal maps behind the scene.

1 Like

What system is used for the GI?
Does it use ray tracing?

According to digital foundry it uses 3 different techniques for diffrent ranges. Voxels for for distant objects, signed distance fields for mid and screen space global illumination for close range. No hardware ray tracing.

What makes it more tangible is that distance fields (altough not used for gi) has been in engine for a long time and ssgi was added in 4.24.

Ssgi worked suprisingly well with light propagation volumes too in Unreal 4, would have been a relatively low cost solution for dynamic gi if lpvs implementation wasnā€™t half finished.

1 Like

The guy here speculates about the tech behind Nanite:
[https://www.reddit.com/r/hardware/comments/gkcd9b/pixels_triangles_whats_the_difference_how_i_think/]

3 Likes

The Tdā€™lr if I understood it correctly is that they are mibmaping the textures and project them in space instead of calculating the geometry and UV and projecting that in spaceā€¦

Todayā€™s gpus are beast ! They can handle billions of polygons easily, the big problem for the PS4 and Xbox was the very weak cpu and the slow storage , but this seems to be fixed in ps5 and Xbox series X
Also the new real time global illumination system itā€™s another mind-blowing achievement , finally a photorealistic image without the need of Ray tracing Gi !!!

Go further beyond,

Letā€™s get some Aiā€™s that take any model and convert them into far smaller procedurals chunks.

1 Like

Tech Analysis: https://www.youtube.com/watch?v=iIDzZJpDlpA

He confirmed itā€™s not geometry texture.
But they use textures for something related to geometry details, also they mentionned 8K textures.

https://twitter.com/BrianKaris/status/1260734555532611584?ref_src=twsrc^tfw|twcamp^tweetembed|twterm^1260734555532611584|twgr^&ref_url=https%3A%2F%2Fs9e.github.io%2Fiframe%2F2%2Ftwitter.min.html1260734555532611584

Also he mentionned they use two asynchronous GPU rasterizers, i think this is also a key to Nanite rendering.

Youā€™ll have to wait for the presentation to know how itā€™s done.