Nvidia Dynamic Multi-Frame Generation Tested: It Works But There Are Issues

Nvidia’s DLSS 4.5 introduces dynamic multi-frame generation to optimize frame rates and latency, showing promise on both high-end and mid-range GPUs but suffering from issues like lack of VSync support, screen tearing, and imperfect responsiveness during rapid scene changes. While improvements such as better UI scaling are noted, the technology remains an early, somewhat flawed implementation requiring manual adjustments and further refinement before it can offer a truly seamless gaming experience.

Nvidia recently released DLSS 4.5 with dynamic multi-frame generation (MFG), a feature designed to optimize frame rate, latency, and image quality by dynamically adjusting frame generation based on performance demands. Tested initially on a high-end RTX 5090 system with Cyberpunk 2077 at 4K resolution, the technology showed promise but revealed significant issues. Notably, traditional VSync is no longer supported, causing frame rates to exceed monitor refresh rates and resulting in screen tearing, which undermines the G-Sync experience. Users must disable VSync manually and set frame rate caps below their monitor’s refresh rate to mitigate tearing, a counterintuitive and inconvenient workaround for what was intended to be a seamless, “fire and forget” solution.

The dynamic frame generation works by increasing or decreasing the number of generated frames to maintain smooth gameplay, but the implementation still feels like an advanced beta. The frame time graphs showed some variability and occasional spikes, indicating that the experience is not perfectly smooth. The RTX 5090’s sheer power required forcing lower DLSS quality modes to trigger dynamic frame generation effectively, while latency remained acceptable, generally around 50 milliseconds. However, the technology’s inability to respect frame rate caps consistently and the resulting tearing highlight fundamental flaws in this first iteration.

Testing on a more mainstream GPU, the RTX 5070, at 1440p showed that dynamic MFG is scalable and can provide a decent experience even on mid-range hardware. The latency was higher compared to the 5090 but still within an acceptable range for many users. However, responsiveness to rapid changes in GPU load, such as looking at different scenes in Cyberpunk’s Phantom Liberty, was imperfect. The system took around 156 milliseconds or more to adjust the frame generation multiplier dynamically, which caused occasional stuttering and overshoot issues during rapid scene changes, reflecting limitations in the granularity and adaptability of the current implementation.

An interesting addition in this release is a new DLSS preset B that improves frame generation for UI elements by recognizing separate UI planes, resulting in better scaling and clarity of interface components during gameplay. Despite these improvements, the overall consensus is that Nvidia’s dynamic MFG is still a work in progress. The technology lacks proper integration with VSync and G-Sync, requires manual tuning of frame rate caps, and exhibits latency and smoothness issues that prevent it from being a fully polished, user-friendly feature at this stage.

Looking ahead, the reviewers suggest that combining dynamic frame generation with dynamic resolution scaling could be a promising future direction, potentially allowing users to balance performance and image quality dynamically. Nvidia’s driver-level awareness of frame times indicates that more sophisticated adaptive solutions are feasible, but they will require further refinement to address current shortcomings. Overall, while DLSS 4.5’s dynamic multi-frame generation is an exciting step forward, it remains an early iteration that needs significant improvements before it can deliver on its full potential as a seamless, high-performance gaming technology.