Glossary
DLSS
Deep Learning Super Sampling — NVIDIA's AI-based upscaler that renders games at a lower internal resolution and reconstructs them to a higher output, recovering performance with minimal visible quality loss.
DLSS uses dedicated Tensor cores on RTX GPUs to upscale lower-resolution renders to a target output. DLSS 4 introduces Multi-Frame Generation, which interpolates additional frames between rendered ones.
Modes
- Quality — internal 67% of output. Usually visually indistinguishable from native.
- Balanced — 58%. Good for 4K with mid-range GPUs.
- Performance — 50%. The biggest fps gain; mild softening in motion.
- Ultra Performance — 33%. Reserved for 8K or budget GPUs at 4K.
DLSS Frame Generation
Synthesizes intermediate frames using motion vectors + optical flow. Doubles or quadruples perceived smoothness but adds latency — best paired with Reflex to keep input lag low.
Competing technologies
FSR (AMD) and XeSS (Intel) provide similar capabilities; FSR 4 closed the quality gap meaningfully on RDNA 4 hardware.
Where this matters
Categories that use dlss
Continue reading