Hacker News Reader

Comment by crote | Hacker News Reader

If you want more detailed models and more complicated scenes, you should really look into raytracing.

The problem with raytracing is that it mostly depends on your resolution, not on your scene. Right now, we're just on the edge of being able to do raytracing at practical resolutions in realtime. But once that's possible, the rest is basically free. Rendering 10M polygons is almost as fast as rendering 1K polygons.

And writing a path tracer isn't that hard - most of the math is quite trivial. You get all the magic shadow / reflection / caustic / lighting stuff for free. We're getting quite close to being able to do movie-level graphics on consumer machines in realtime, and all that without any of the tricks required for okay-ish looking rasterizer shading. Best of all, it's even easier to work with for artists as well!

View on HN

"Rendering 10M polygons is almost as fast as rendering 1K polygons"

That is correct for ray tracing.

But modeling and animating 10M polygons requires a Hollywood movie budget, while I can model and convincingly animate 1K polygons myself.

Of course, but at least now you have the ability to render it. "Black Ops 2" made $1B in 15 days, so some studios already have this budget, so they should be able to make use of it.

As for smaller players? Personally, I'd be interesting in procedural generation, or just modelling NURBS surfaces. And have you seen the demos of raytraced Minecraft / Doom? The highly advanced lighting already makes it look way more detailled, even with a low poly count. Even low budget games will be able to make good-looking games.

The Doom+Raytracing pictures look great, but I believe they are fake

https://www.dsogaming.com/news/doom-rtx-released-mod-bringin...

For Minecraft, the water is a bit nicer, but it still looks very unconvincing to me:

https://www.youtube.com/watch?v=nXoLY9lF-NI

Unless I'm missing something it's not a very fair comparison, you have low-res unshaded textures on one side and high-res bump-mapped textures on the other. That alone accounts for 99% of the difference between both versions.

Meanwhile current "AAA" games that have RTX support don't look radically different with RTX on or off. Better and more accurate reflections of course but also a big hit on performance that's, IMO, not worth it.

Of course that may just be because the devs don't bother to use the tech to its full potential when only a tiny fraction of the playerbase will be able to use that mode anyway.

High quality models are created via sculpting (5-10 million is actually still on the low end for that) before being retopologized. For anything high-fidelty (that is, anything you might encounter in a non-hobbyist project) it's _more_ work to create a model with a low polycount. The human work amount for animation is also mostly dependent on the rig, so if you're fine with a simple rig, animating a high polycount mesh isn't really any harder.

What kills it though is that even though raytracing scales well, animation/simulation does not.

I'd guess that rigging a 10 mio mesh will be pretty much impossible in 3ds max / Maya / Blender. Their viewports are not built to handle such an insane polycount.

Plus, I wouldn't know of any UV-mapping tool that can handle a 10 mio poly count.

Well, you could also use ZBrush for rigging. I think that multi res in Blender can work with up to 10 million polys (or maybe this is a work in progress?), but don't quote me on that.

I don't see why you wouldn't just go for vertex colors instead of UVs. Polypainting works fine even at 10 million works in all tools I've worked with (zbrush, substance painter, 3dcoat). If there's some reason you'd want UVs, you could just bake them to UVs afterward.

>But modeling and animating 10M polygons requires a Hollywood movie budget, while I can model and convincingly animate 1K polygons myself.

I can make 10M polygons just by opening blender and asking for a UV sphere with a bunch of sides... Artists using the sculpt tool can produce models with humongous numbers of polygons, and then they have to work really hard to optimize them. I would argue the exact opposite of your claim. Models with more polygons are easier to end up with than models with fewer.

I think the argument here is that creating quality assets with those levels of detail is expensive, not that you can create large and unwieldy meshes by not understanding how to model efficiently... that goes without saying.

"model efficiently" implies optimizing a benefit for a cost. In this case, I think the nature of the cost has completely changed. The definition of efficiency will probably have to change also.

Not really.

A 10 mio poly mesh with a typical XYZW+RGBA+UV+UV2 = 12 float stream format would require 12x4x10 mio = 480 MB of GPU memory per copy. Skinning needs weights (200 MB) and at least a second copy, so one single 10 mio poly mesh would eat up about 1.3 GB of GPU memory.

So even with ray-traycing, optimizing models for a manageable polycount will remain crucially important.

Reply on news.ycombinator.com