NVIDIA has released the latest generation of graphics cards, and I have acquired a unit for use as part of the Proof of Concepts I am conducting.
I'll be using this card together with the external GPU housing to provide a portable environment for inferencing (running object detection on top of real time video). It provides significantly more computational power than our current consumer grade cards.
Typically past generations of NVIDIA graphics cards have gone up in increments of 100- the GTX780, GTX980, GTX1080 etc. The ability to do real time ray tracing has led NVIDIA to completely "jump a generation", with the new RTX 2080Ti (RTX representing the ray tracing capability).
Ray tracing is a rendering technique that is able to produce incredibly realistic lighting. It simulates rays of light, and how it interacts with other objects. Imaging covering a light with your finger- you will see a diffuse red glow as the light shines through your skin. Similary, ray tracing provides significantly improved shadows, reflections, translucence and scattering of light.
Ray tracing has been used in movies for years, however each individual frame may have taken upwards of 90 minutes to render. Now, ray tracing is possible in real time, at speeds of 60-70 frames per second.
The two videos below show what light looks like when rendered through ray tracing, and without it.
Deep Learning Super Sampling
Aliasing is the distracting jagged edges of objects that are being rendered on a screen. Traditional temporal anti-aliasing techniques are very computationally heavy, and thus NVIDIA have released a neural network capable of exceeding the performance of traditional techniques whilst providing a crisper image.
The full Battlefield V video game trailer below showcases the impact of ray tracing. Time will tell whether support from the video games industry will help this technology flourish, or fade into obscurity. Here's hoping for the former!