Download

Tesselation needs to come back, desperately

This is nonsense.

Sure, Nanite is great, but in it’s current shape and form absolutely NO replacement for tesselation.

So, with a bit of testing, I found the following issues due to the lack of tesselation.

  • There is NO way to subdivide a mesh on the fly. So no taking a mesh from, say, the starter contents, and throwing tesselation on it so you can subdivide it quickly. Now, the only way to do it is to first enable modelling tools from the plugins, create a new mesh, subdivide that mesh and then save it. Even so, that will results in massive polygon counts that will slow down the scene. Wasn’t the whole point of the updates to make the development process easier? Then why would you just introduce a crapton of steps we didn’t have to deal with before?
  • Enabling Nanite makes it so that you cannot use WPO; I tested it and a shader with WPO on a nanite enabled mesh didn’t render. Turning either Nanite or WPO off would render it. Seriously, what the hell? You expect us to just forego ANY vertex animated meshes except for the most rudimentary? Because that is very poorly thought out. What about giant ocean shaders? What about other special effects that require high polygon count mesh animations? Sure, Nanite is impressive. But with these sort of limitations, don’t go deprecating technology that actually has features Nanite doesn’t have. Some of us actually use that still. There is more to life than photoscanned assets, you know.

This sucks. Unreal 5 looks amazing, but these sort of limitations put heavy dampers on what Unreal 4 is already perfectly capable of doing. It’s like it took 1 step forward and 2 steps back.

So put tesselation back please. At least until Nanite can at least do what tesselation already could.

6 Likes

Unreal has its own water surface now that handles distance based tessellation

Don’t use nanite for those? You can mix nanite meshes and non-nanite meshes in the same scene.

3 Likes

Unreal has its own water surface now that handles distance based tessellation

Great, I did find those tools and they seems to be pretty nice, even if they are hidden behind a plugin that is not enabled by default. Still, why is that distance based tesselation not a feature in materials? Do we have to program a solution by ourselves now? It also seems, again, highly counterintuitive. That nifty looking WaterFluidSimTest map? I cannot seem to change the mesh. Admittedly though, the other water assets seem pretty promising.

Don’t use nanite for those? You can mix nanite meshes and non-nanite meshes in the same scene.

That is literally not solving any of the entire issue I mention. This does not solve the lack of ability to subdivide a mesh on the fly. I now have to bring out the Unreal tools, or go back to the 3D software, export a high polygon mesh, put it into the game. It also doesn’t solve the fact that I will have to deal with this high polygon mesh eating up performance whenever it’s visible. You know, issues I didn’t use to have when I could actually use tesselation. And sure, tesselation is not without its performance issues either, but I would rather have the choice, especially if Nanite doesn’t even support WPO animation to begin with.

1 Like

I agree this is not terribly convenient, but it only takes a few minutes to pre-tessellate your mesh and apply displacement to it (in ZBrush). I can understand that tessellation is convenient to use but considering it is a huge step backward in quality it seems like the solution should be to make nanite more convenient, not to go back to tessellation.

Just my opinion.

Not entirely sure what you mean by this. The whole point of Nanite is that it is (fairly) fixed cost and doesn’t scale with scene/mesh complexity. It’s why Epic recommends it for virtually everything it can be used for, whether you’re using millions of triangles or not.

4 Likes

Adding deformation to Nanite has to be one of the most requested features. I have no idea how complicated it might be or if it’s even possible, but if they do that, we could have stuff like fields of modeled vegetation reacting to wind or the most absurdly detailed character we could imagine.

1 Like

It’s important to remind ourselves this is also early access software and still a long way from release quality. Complaining a lot at this stage about features that may change or are very likely to change and be improved is moot at best. You should expect that the UE5-EA is not going to provide what you need. It’s really very early and not feature complete. If UE4.26.2 has the features you require then use that. It’s not progressive in software development to drag forward a legacy feature like tesselation when there are solutions being built to make it obsolete, even if those solutions may not be complete at this time. Give the software time to mature. It’s very early still.

7 Likes

If they can make nanite capable of at least dealing with WPO, tesselation can go for all I care. That is my primary gripe; that Nanite doesn’t render with WPO enabled. If they can get that kink worked out before the actual UE5 releases, my complaint wijll basically be moot, but if not, there is still value in having tesselation in there.

What I mean by high polygon mesh eating up performance was the high polygon mesh without nanite enabled of course. After all, if we want to enable WPO, we cannot have nanite enabled at the same time at this moment. Basically, it would require us to just have to put the high-poly model in there wholesale, without any nanite optimization. Sure, with tesselation we would move that required performance to the GPU instead, but it was an option at least.

It’s important to remind ourselves this is also early access software and still a long way from release quality. Complaining a lot at this stage about features that may change or are very likely to change and be improved is moot at best. You should expect that the UE5-EA is not going to provide what you need. It’s really very early and not feature complete. If UE4.26.2 has the features you require then use that. It’s not progressive in software development to drag forward a legacy feature like tesselation when there are solutions being built to make it obsolete, even if those solutions may not be complete at this time. Give the software time to mature. It’s very early still.

Yes, true, I am completely aware that this is a early access. Which is why I am posting here, as I hope that with me griping about it, I may add to the ever increasing pile of people who complain about it, thus perhaps showing Epic they may need to do something about it. So that by the time the actual for-production version comes out, we have something that works with WPO.

Yes while this is early access i do remember how long we were waiting for water shaders in ue4 or anything that became unusable in screen space shader rendering that was easy to do in older versions like UT2007 or UDK.

So complaining now (even if this is early access) makes sense.

4 Likes

Ow man, I remember that. We had this super nifty rendertarget interaction thing from UDK (at the time anyway), which became completely unusable in UE4. Sure, we had very little access to that functionality (in that we could only make a water surface with some basic interactives, and that was about it), but I missed that when I started using UE4.

The UE4 documentation tells me it took them until 4.13 to finally put that back in. I really hope WPO in Nanite implementation will be a bit faster. Or that we at least get something of a solution instead before 5.13 or something.

1 Like

I woinder how they plan to use landscape displacement/tesselation in the future… i can’t imagine them removing that…

Nanite will support WPO in the full release, don’t get your pantines in a twist. Read the docs / release notes before raging on forums.

3 Likes

water looks pretty tesselated to me
water ue5 - YouTube

It looks like it, but it isn’t. It’s a custom meshing system designed specifically for the water, similar to how Landscape LODs work.

1 Like

alright, manual tessellation ftw :wink:

Final release won’t include tesselation either world position offset

Another aspect of this is that pre-tessellating your models comes with larger file sizes for no good reason, and that’s only if your use case is covered by Nanite at all.

If hardware tessellation as we know it is no longer used but existing UE4 materials import and do, e.g., something like the water, I think that’s OK.

We are likely speaking about minor movements, so that cluster bounds can be inflated and account for that during culling. Arbitrary WPO is unlikely to happen with Nanite, only relatively small motion.

As for tess, how about just mesh shaders? Tess is pretty dated and is becoming extinct. I’ve heard a lot of mentions that tess is very much need, but can anyone at least give an example of UE4 product, that makes best use of tessellation? While there might be, I don’t remember noticing it making visual impact in majority of recent UE games.

The first thing that comes to mind is things like snow build up and other similar effects. Maybe a character gets a similar effect on them. I haven’t played around it for like 10 versions because I’ve been focused on technical things. It seems like the suggestion is just to pre-tesselate everything with UE5 or use specialized systems.

I think most of the times tesselation came up in games was with skeletal meshes until Crysis 2 happened.

The only major game I can think of which used tessellation in a meaningful way in recent memory is the Demon’s Souls remake. Of course that’s not UE4 so…

As for WPO, they talked about general deformations on the Nanite stream and how WPO falls under that together with skinning etc. and they mentioned that this is something they’re looking to support, so I suppose we’ll see what happens.