I think calculating the rays in a different way does constitute as rethinking the pipeline, especially when we consider that path tracing is one of the most computationally heavy processes in computer graphics. In fact path tracing is so heavy we don’t even do full path tracing (as in we don’t calculate all the possible rays), we essentially cheat by calculating a handful of rays and then sending it through a denoiser (which is why it takes a second to calculate the shadow of your character). There’s a lot of performance to be found in raytracing and if they’ve found some then that’s a pretty big deal.
I do know all of this, it’s just dedicated hardware for a step we’re currently simulating in shaders. Dedicated hardware that if I’m not mistaken exists on NVIDIA graphics cards already.
That’s an added capability, not a rethinking. But it will enable raytracing in a way that is far less expensive.
Yeah, not sure if “use upscaling, raytracing and compress a little more” counts as “rethinking”
I think calculating the rays in a different way does constitute as rethinking the pipeline, especially when we consider that path tracing is one of the most computationally heavy processes in computer graphics. In fact path tracing is so heavy we don’t even do full path tracing (as in we don’t calculate all the possible rays), we essentially cheat by calculating a handful of rays and then sending it through a denoiser (which is why it takes a second to calculate the shadow of your character). There’s a lot of performance to be found in raytracing and if they’ve found some then that’s a pretty big deal.
I do know all of this, it’s just dedicated hardware for a step we’re currently simulating in shaders. Dedicated hardware that if I’m not mistaken exists on NVIDIA graphics cards already.
That’s an added capability, not a rethinking. But it will enable raytracing in a way that is far less expensive.