AI Audit: Nvidia DLSS
About This Project
This website was created for our Critical AI Studies class where we selected an AI application to analyze in depth. Our team chose NVIDIA's Deep Learning Super Sampling (DLSS), an AI-powered technology that improves game performance by upscaling lower resolution frames.
Analysis
History
The evolution of DLSS from version 1.0 to current day
Rhetoric
Analyzing NVIDIA's marketing language and how consumers view the product
Data
Examining what data is used to train DLSS models and how it's processed
Materials & Labor
Investigating the physical hardware and work behind DLSS
Outputs & Consequences
The impact of DLSS and graphics cards' effect on the world
Understanding DLSS
What is DLSS?
DLSS allows games to render at lower resolutions to improve performance, then it upscales the image to enhance the image quality.
DLSS Super Resolution
- Renders games at lower resolutions to boost performance
- Uses AI to upscale to higher resolutions
DLSS Frame Generation
- Creates entirely new frames between rendered ones
- DLSS 3 generates one frame between each rendered frame
- DLSS 4 can generate up to three frames per rendered frame
How It Works
NVIDIA trains DLSS on supercomputers by comparing its output against ultra-high quality 16K reference images thousands of times.
Advanced Features
- Ray Reconstruction: Improves the quality of ray-traced scenes
- Deep Learning Anti-Aliasing (DLAA): Uses DLSS technology to improve anti-aliasing at native resolution
Native vs DLSS
Native Rendering (left) vs DLSS (right)
Performance Comparison
Note: while this chart claims an 8x "performance" boost, these results include frame generation, which increases frame rate at the cost of pc latency.
DLSS Detection Challenge
Can You Spot the Difference?
Test Your Eye
Can you tell which images use DLSS upscaling and which are rendered at native resolution?
Question 1 of 5
Which of these images uses DLSS upscaling?
