NVIDIA Announces Enhanced DLSS Frame Generation for RTX 40 GPUs: Benefits Include Reduced VRAM Use and Improved Performance

NVIDIA Enhances Frame Generation and Reduces VRAM Consumption on RTX 40 GPUs

NVIDIA is set to improve its GeForce RTX 40 series GPUs with a new artificial intelligence model designed to optimize Deep Learning Super Sampling (DLSS). This update aims to boost frame generation efficiency while minimizing the VRAM required for enhanced gaming performance.

Initially introduced with DLSS 3.0, the Frame Generation feature has been continually refined, significantly improving upscaling quality and gaming performance. This feature is particularly valuable as it helps fill the gaps between frames, resulting in a smoother gameplay experience.

Recently, NVIDIA revealed DLSS 4.0 in conjunction with its next-gen RTX 50 series GPUs. While this upgrade will be compatible with the RTX 40 series, it will not include the multi-frame generation capability present in the 50 series. Nevertheless, NVIDIA is committed to enhancing the frame generation technology for RTX 40 users too. According to Jacob Freeman from NVIDIA, both RTX 40 and RTX 50 series GPUs will benefit from upgraded frame generation.

The forthcoming updates promise to deploy a new AI model that will not only enhance performance but also lower latency and reduce VRAM consumption by up to 400 MB in specific scenarios. This improvement allows the RTX 40 GPUs to allocate more resources to other critical game elements, such as textures and mods, which often require higher VRAM availability.

In practical terms, even minor enhancements like the 400 MB reduction in VRAM usage can significantly influence gaming experiences, particularly for users with 8 GB GPUs. In addition to refined frame generation, DLSS 4.0 introduces features such as Enhanced Ray Reconstruction, Super Resolution, and Deep Learning Anti-Aliasing (DLAA) for the RTX 40 series.

NVIDIA DLSS 4.0 Features — Image Credit: nvidia.com

DLAA is currently in beta and is compatible across all RTX GPUs from the 20 series to the 50 series. Notably, while frame generation is limited to the RTX 40 and 50 series, RTX 40 GPU owners can still expect measurable performance boosts, despite missing the advanced multi-frame generation capabilities of the latest GPUs.

However, an important area of focus will be input latency. The introduction of DLSS 3.0 initially resulted in increased latency due to frame generation processes. NVIDIA asserts that DLSS 4.0 will address this issue, although generating additional frames typically requires more time for prediction and rendering.

NVIDIA’s comparisons indicate minimal differences in latency between DLSS 2.0 and the newer DLSS versions, 3.5 and 4.0. While it’s possible that Reflex was disabled during DLSS 2.0 testing, the inclusion of Reflex in the latter two versions suggests that latency experiences in DLSS 4.0 are comparable to those of DLSS 2.0, which raises questions regarding optimization.

Source & Images