SIGGRAPH 2018 – Nvidia Delivers Ray Tracing On A Card

I had the opportunity to be at #SIGGRAPH this year and it was a good year to be here. Special Interest Group on Graphics (SIGGRAPH) is the seminal computer graphics conference that jump started the computer graphics industry with early luminaries from both Pixar and academia. The roots span more than three decades of ground-breaking computer graphics technologies advancing the art and technology that comes together to form the modern media and entertainment industry.

Not to burst Nvidia’s bubble, but as I’m wearing an Nvidia lanyard around my neck and I would be remiss not mention Jensen Huang’s keynote presentation. Nvidia is launching a Quadro RTX, a ray-tracing card that uses the latest Turing architecture and combines ten years worth of research of Artificial Intelligence and Machine Learning with a new RTcore (ray tracing core). Ray-tracing has long been gold standard for producing photo realistic content. Pick up any basic optics book since Opticks by Issac Newton and there are methods given for representing light as rays called the ray model of light with formulae for reflection, refraction, and such. Newton had described corpuscles, essentially particles, which followed geometric rules and ratios. Without consideration of the quantum world, it works well as a model for the light and behavior we see.

The tipping point for this next generation of card is really the ability to render real-time on a single graphics card made possible by combining the ray tracing with machine learning techniques. In the most basic algorithm for ray tracing, every incident ray of light that strikes a surface generates additional rays of light which in, turn produce more rays of light. This by itself generates exponential cycles to compute. Instead, machine learning is used after a certain threshold which means Artificial Intelligence can be used to fill in the remaining details. Jensen claimed this brought traditional ray tracing solution from 5-10 years out to cards available today. All of which adds up to be a pretty exciting time to be working on graphics and machine learning.

Tags: , ,

Leave a Reply

Search

Websites

  • slide
  • slide
  • slide
  • slide