In this video, Andrew Edelsten, a Deep Learning Engineering Director at NVIDIA, details the three most important things developers need to know about DLSS.
You can learn more by attending Andrew’s talk at GDC:
Truly Next-Gen: Adding Deep Learning to Games & Graphics
Presenters:
Andrew Edelsten (Director, Developer Technologies (Deep Learning), NVIDIA)
Paula Jukarainen (Developer Technology Engineer, NVIDIA)
Anjul Patney (Senior Research Scientist, NVIDIA)
Location: Room 205, South Hall
Date: Thursday, March 21
Time: 12:45pm – 1:45pm
Pass Type: All Access, GDC Conference + Summits, GDC Conference, GDC Summits, Expo Plus, Audio Conference + Tutorial, Indie Games Summit
Topic: Programming
Format: Sponsored Session
3: DLSS (Deep Learning Super Sampling) uses the power of AI to boost frame rates in games with graphically-intensive workloads
For instance, when we look at the next-gen visuals in 3D Mark’s “Port Royal”, we can see that even NVIDIA RTX GPUs with their RT Cores can struggle with the benchmark’s extreme content. By enabling DLSS, which leverages Turing’s Tensor Cores, we have a much higher and more consistent frame rate.
2: DLSS Mapped a Fast Path from Deep Learning Research to Product
The NVIDIA Research team invented DLSS, but adding it to a game required extreme optimization of the model and experimenting with different layer types, input sizes, and training techniques. Going from research to production was a big step, during which we learned a ton and added layers of automation to our Saturn 5 Supercomputing Cluster.
1: It’s the First Time We Have Put Real-Time Deep Learning Right in a Game
This is just the tip of the spear! Over the next few years, we may see games with:
- Text-to-Speech
- Speech-to-Text
- Actual “AI” for Bots
- DL-Driven Dynamic Materials
- DL based Animation Systems
- And 50 Things We Haven’t Even Considered Yet
It’s going to be awesome! So start thinking today about how you’ll use deep learning in your next game!
If you’re attending GDC 2019, come and learn more about DLSS, how DLSS was integrated into Battlefield V, and Deep Learning for games and graphics, during the session listed at the top.
