Looks pretty bad for now.
The Developer usually is very talented and creative. Made this for example:
Really a remarkable game, liked it.
Looks pretty bad for now.
The Developer usually is very talented and creative. Made this for example:
Really a remarkable game, liked it.
I dont see anything that couldnt be done in 4.Its only on pc for now, yes. No mention of a ps5 port as far i can see.
Looks stunning when you actually play it, quite different from your usual game (using light and darkness to gain advantages).
I dont see anything that couldnt be done in 4.
Looks underwhelming compared to the tech demos we saw on U5
Slides from Nanite deep dive at SIGGRAPH are posted here: http://advances.realtimerendering.com/s2021/index.html
Hellblade 2 was in development before ue5 early access so its featureset should mainly focus on things possible before. It might run on ue5 but it is mainly an ue4 port so far.Love what they can do with Metahuman. Can't wait to see all future UE5 projects especially Hellblade II.
Ninja Theory mentioned at E3 that they are still working on a vertical slice of HB2.Hellblade 2 was in development before ue5 early access so its featureset should mainly focus on things possible before. It might run on ue5 but it is mainly an ue4 port so far.
Yeah unless otherwise stated it's usually PS5.We probably don’t want apis to enforce cluster or instance formats but a hardware micropoly rasterizer would be useful. This could help with Nanite’s overdraw problem though performance seems fine already. Presumably the stated performance numbers are from PS5.
Conceptually the clustering part (and even the simplification part) can work for skinned stuff that does not change topology. Some sort of cluster culling is going to be desirable for stuff like virtual shadow maps regardless of Nanite. The difficulties arise more around if arbitrary "late" deformation is allowed you still need to understand bounding volumes at a coarse scale, ideally without skinning all the vertices in memory first. Obviously that can work when geometry isn't too high poly (indeed that is how people are forced to do it for raytracing), but would not be feasible for the amount of detail that the Nanite meshes in the demos typically have. For those I imagine you really want to skin some sort of control mesh instead.The problem with deformable meshes is only relevant when you need multiple LODs right? Is there any reason Nanite couldn’t be used for 3rd person character models or first person weapon models where you always want maximum LOD?