This looks very impressive. 60 volumetric cloud layers, plus 20 data layers of various other atmospheric conditions. Sounds like they're utilizing RTX to fullest because I doubt this is possible for real-time without ray-tracing acceleration hardware?
But this part in the article sounds weird to me:
the new engine brings higher-fidelity collision detection, which makes things like bumpy runways, braking on slick surfaces, varying degrees of friction, etc. much more nuanced. Under the hood, the development team has done significant work with integration, using something called adaptive time steps. The upshot is that the simulation now always updates at the same frequency as the visual frame rate.
This sounds more like the physics solver can run adaptively at whatever timestep currently needed, based on the simulation conditions for the frame, no?
Edit: Sounds like a framedrop results in a dropped physics solver iteration as well I suppose?
It was in development for 5 years so they had to go with existing technology, there were no predictions of RTX hitting consumer market in 5 years. Even then RTX is still early adopters stage, slow, buggy and expensive.
They said that they aren't using raytracing I believe. That'd be great if they did, I don't think there are any killer apps for a 2080ti yet, but this would certainly be one.
I don't think there's anything here that would particularly benefit from raytracing. Raytracing usually shines when modeling stuff like reflections, refractions, and multiple light sources that are expensive with shaders, but I don't see much of that here.
this extremely interesting video show state of the art techniques used to fake global illumination without using ray-tracing and compares the result with the ray-traced variant https://youtu.be/CuoER1DwYLY?t=127
long story short moving object color casting light is the main visible difference that cannot be replicated with traditional techniques (like baking the light reflection in the scene at built time, but can be faked convincingly); ray-tracing is also not limited at reflecting elements that are on the scene like what you'd get from screen space reflection, which is less noticeable of an effect but is there if you pay attention and know what to look for.
Agree, but I suppose I meant more in terms of optimizing performance by making use of idle 2080TI RT cores. Volumetric cloud lighting as OP said, cabin reflections, water perhaps..
I'll take anything that would mean the community doesn't need to keep modifying .CFG parameters to get an extra few FPS
Comments
This looks very impressive. 60 volumetric cloud layers, plus 20 data layers of various other atmospheric conditions. Sounds like they're utilizing RTX to fullest because I doubt this is possible for real-time without ray-tracing acceleration hardware?
But this part in the article sounds weird to me:
This sounds more like the physics solver can run adaptively at whatever timestep currently needed, based on the simulation conditions for the frame, no?
Edit: Sounds like a framedrop results in a dropped physics solver iteration as well I suppose?
Nope, surprisingly it is a DX11 application with a traditional rendering pipeline. This is all PBR-rasterized; nothing raytraced.
They said they might have plans to port to DX12 in the future, but not before the release.
It was in development for 5 years so they had to go with existing technology, there were no predictions of RTX hitting consumer market in 5 years. Even then RTX is still early adopters stage, slow, buggy and expensive.
DX12 is older than 5 years so I'm still a bit surprised on that part. Porting to DX12 is not easy. it's like OpenGL vs Vulkan. Very different APIs
They said that they aren't using raytracing I believe. That'd be great if they did, I don't think there are any killer apps for a 2080ti yet, but this would certainly be one.
I don't think there's anything here that would particularly benefit from raytracing. Raytracing usually shines when modeling stuff like reflections, refractions, and multiple light sources that are expensive with shaders, but I don't see much of that here.
I thought rendering layered volumetric stuff is where RTX would shine but I don't know much about that type of rendering tbh.
this extremely interesting video show state of the art techniques used to fake global illumination without using ray-tracing and compares the result with the ray-traced variant https://youtu.be/CuoER1DwYLY?t=127
long story short moving object color casting light is the main visible difference that cannot be replicated with traditional techniques (like baking the light reflection in the scene at built time, but can be faked convincingly); ray-tracing is also not limited at reflecting elements that are on the scene like what you'd get from screen space reflection, which is less noticeable of an effect but is there if you pay attention and know what to look for.
Agree, but I suppose I meant more in terms of optimizing performance by making use of idle 2080TI RT cores. Volumetric cloud lighting as OP said, cabin reflections, water perhaps..
I'll take anything that would mean the community doesn't need to keep modifying .CFG parameters to get an extra few FPS
For my RTX2080 (non ti) the killer app is VR, all the bells and whistles to the max and 80fps solid (Rift S) is worth it.
Outside of that use case a 2070 super makes far more sense (wasnt out when I bought mine though).