The idea is that since there are several years ahead until full raytracing hardware starts to appear, you could focus on that (and a single Vulkan renderer), reducing the manpower needed.
Meanwhile, commercial engines have to focus on current games, rasterization, raster + RT and the myriad of renderers and hardware they support.
Full raytracing is very possible sooner than we may think if approaches like DLSS 2 keep improving.
I thought you were saying they should focus on ray tracing instead of the Vulkan renderer. Saying they should build a vulkan renderer and a ray tracing renderer is a much different statement than what I was understanding, and sounds like a lot to ask for.
Additionally there's a ton of risk in betting on full ray tracing in several years. Given that they are a small team with a small community and small funding, they simply have to hedge there bets, and myself as a user am damn glad they are. Unity going guns blazing on new shit and having things be broken all the time is a huge reason why I switched to Godot in the first place.
A nice thing about godot is that it's code base is very easy to get your hands in to, so if Juan wants to focus on the big stable market with the raster renderer, the community can work on a ray tracing renderer. Juan is exposing an api called RenderingDevice in 4.0 that should enable the community to write their own renderers easier than from scratch.
> Saying they should build a vulkan renderer and a ray tracing renderer is a much different statement than what I was understanding
Raytracing is an extension on Vulkan. I don't understand what you are trying to say. I am not talking about CPU raytracers.
> Additionally there's a ton of risk in betting on full ray tracing in several years.
They are not commercial, so the risk is minimal. That is why Godot is such a good place for that.
> Unity going guns blazing on new shit and having things be broken all the time is a huge reason why I switched to Godot in the first place.
That is true and it is why most games avoid updating Unity if possible. But I feel if Godot had so many features, it would have the same problem too (or worse, given less manpower).
> the community can work on a ray tracing renderer
Juan is the one with financial support. Most people cannot afford to take on a research-heavy, multi-year project on their own.
Nevertheless, RenderingDevice sounds great for research projects and academia!
Not technically wrong, but an incomplete picture. The way that engine renderers are organized right now does not lend them well to raytracing. The Vulkan extension we're talking about requires you to set up completely separate pipelines, use different kinds of dispatches, and different shaders. Very little is compatible, and most "ray-traced" are bodged in there in a weird way -- they're mostly used for single-bounce specular AA, as we only have a very small ray budget.
I was pointing that out because the other user seemed to think I was talking about a non-accelerated renderer. However, to access RT hardware and compute capabilities you need Vulkan nevertheless (or the non-portable ways that I guess Godot wants to avoid).
My original post was precisely about focusing on such a full RT renderer (via Vulkan). I am not an expert, so it may be a dumb idea, but it would be nice to have a FOSS production-quality renderer by the time full RT becomes a thing for more and more games.
Betting your project development on something that might pay of in the future is not minimal risk, both for the project and for the people being financially supported to work on.
I bought a new system with RTX 2070 Super and couldn't get any demos with raytracing to work acceptably in either Unreal or unity... The frame rates on super simple levels were maybe 10fps at best...
I was a little dismayed. It seems like AAA games are also not making a huge amount of use of it either. Despite the RTX name I don't know if it's really ready for prime time, at all.
I have a 2070 non super. Minecraft RTX was acceptable. You can't get competitive FPS gamer acceptable framerates but you can get RPG/minecraft acceptable framerates (40ish) depending on how spoiled you are by 144+ fps feels.
But yeah generally I agree, a game that is already doing AAA amounts of gemeometry and shading, that uses RTX to good effect, is going to need 2080ti in SLI to be still a bit slow.
Current gen hardware is nowhere powerful enough for that, yeah. But for simple games a lot is already possible (like Minecraft).
DLSS 2 and similar techniques help a lot to reduce the number of rays needed. Perhaps by the time of RTX 4000 (what I meant as next-next-gen) we might be able to see full raytracing.
I have an RTX 2060. I was very impressed by the game Control. It uses raytacing and has good frame rates. It's graphics are really nice. It certainly looks possible to get good results with current hardware, particularly now DLSS 2 is as available.
Considering they don't have enough manpower to compete with UE/Unity in raster, would it make sense going for full raytracing now?
That way, when next-next-gen hardware arrives in 2-3 years, they could be on the forefront of editors specialized in fully raytraced games.