Nvidia’s Deep Learning Super Sampling (DLSS) has reached version 4.0, marking a significant leap in AI-powered graphics enhancement. DLSS 4 introduces groundbreaking features like multi-frame generation, alongside improvements to upscaling and a new real-time “transformer” architecture similar to that used in large language models like ChatGPT. This article explores the key advancements in DLSS 4 and their impact on gaming performance.
Multi-Frame Generation: A Game Changer
DLSS 4’s standout feature is multi-frame generation. This AI technology generates up to three additional frames from a single rendered frame, significantly boosting performance. This is how Nvidia’s RTX 5090 achieves impressive frame rates exceeding 200 fps in demanding titles like Cyberpunk 2077 and Black Myth Wukong, even at 4K resolution with ray tracing enabled.
Multi-frame generation requires a minimum of 60 fps, even with DLSS upscaling, to avoid visual artifacts. While it can sometimes produce ghosting or a motion blur effect, especially at the 4X setting, the 2X and 3X options offer substantial performance gains. The increased input latency introduced by frame generation is mitigated by Nvidia’s Reflex 2 technology.
Currently, multi-frame generation is exclusive to the RTX 50-series GPUs, like the RTX 5090 and 5080. While there’s speculation about potential optimization for older RTX 40 and 30-series cards, no official confirmation exists.
Enhanced Super Resolution: Sharper Visuals for All RTX GPUs
The core of DLSS remains its Super Resolution upscaling, allowing games to render at lower resolutions and then upscale the output using AI. DLSS 4 significantly improves this core functionality, benefiting all RTX GPUs, from the 20-series to the latest 50-series.
The new transformer model analyzes more parameters from the rendered frame than previous Convolutional Neural Networks (CNNs), resulting in superior image quality. This translates to reduced ghosting and artifacts, improved motion resolution, and smoother edges. The transformer model also enhances other RTX technologies like Ray Reconstruction, making ray tracing more efficient and visually appealing.
DLSS 4 Game Support and Overrides
Unlike previous DLSS versions that faced initial limited adoption, DLSS 4 launched with support for over 75 games, including titles like Indiana Jones and the Great Circle, STALKER 2, Alan Wake 2, Black Myth Wukong, and Cyberpunk 2077.
For games without native DLSS 4 support, Nvidia now provides DLSS overrides through the Nvidia app. This allows users to force the transformer model and enable quality settings like Ultra Performance and DLAA, even if not officially supported by the game. You can find the full list of supported games here.
Alternatives to DLSS 4
If you don’t own an Nvidia RTX GPU, alternatives like AMD’s FidelityFX Super Resolution (FSR) and Intel’s Xe Super Sampling (XeSS) offer upscaling capabilities in compatible games. For frame generation, while Nvidia’s RTX implementation remains exclusive to its hardware, Lossless Scaling, a third-party application available on Steam, offers similar functionality for a wider range of hardware.
Conclusion
DLSS 4 represents a substantial advancement in AI-driven graphics technology. The introduction of multi-frame generation and the improved Super Resolution powered by the transformer model deliver significant performance and visual enhancements. While certain features remain exclusive to the latest RTX 50-series GPUs, the advancements in Super Resolution benefit all RTX card owners, promising a more immersive and visually stunning gaming experience.
