Nvidia officially announced a significant update to its Deep Learning Super Sampling technology for RTX 50-series owners, introducing Dynamic Multi Frame Generation. This new feature allows graphics cards to switch automatically between frame generation modes to maintain optimal performance on high refresh rate displays. According to PC Gamer, the update aims to balance visual smoothness with system responsiveness across various gaming scenarios. The technology was first previewed at CES in January before becoming available to users for testing.
Technical Mechanics of MFG
The updated system enables the graphics processing unit to interpolate between rendered frames using advanced artificial intelligence models. Instead of locking into a single multiplier like 2x or 4x, the software now selects modes ranging from 2x up to 6x dynamically. This adaptation relies on the monitor maximum refresh rate and real-time performance metrics to determine the best scaling factor. PC Gamer explained that the multi part in MFG refers to the fact that the generative stage can generate multiple frames.
Nvidia also rolled out a second AI model known as Preset B to improve the clarity of static user interface elements. This specific enhancement requires games to expose a UI depth buffer, which limits its applicability to approximately 20 currently supported titles. > "The new version of DLSS MFG also includes an updated AI model that enhances in-game user interfaces," Nvidia stated in a press release. This model cannot provide a benefit to games that do not expose the necessary depth buffer data.
Reviewers tested the technology using a Ryzen 9 9950X3D processor paired with a GeForce RTX 5090 graphics card. The display hardware included an MSI MPG 321URX 240 Hz OLED monitor to evaluate high refresh rate capabilities. System latency and frame pacing were measured using Nvidia App statistics overlays during the benchmarking process. Reflex settings were disabled for baseline runs to establish a clear picture of native performance.
Game Performance Analysis
In Cyberpunk 2077, the dynamic system successfully switched between 4x and 5x modes to approach the target 240 fps. While the frame generation reduced jankiness, it increased system latency compared to native rendering without generated frames. Tests showed the PCL figure rising to around 50 milliseconds when the system engaged higher generation modes. The game still looked janky without frame gen, but the dynamic system managed to smooth out the pacing significantly.
Dragon Age: The Veilguard demonstrated different behavior, as the game ran smoothly without needing aggressive frame generation. Dynamic Multi Frame Generation remained in 2x mode throughout the test, preserving lower latency while achieving near 240 fps. This stability suggests the algorithm avoids unnecessary scaling when native performance is sufficient for the display refresh rate. The system latency was actually lower with frame gen enabled due to Reflex synchronization mechanisms.
Hogwarts Legacy showed significant variance in frame rates, particularly in areas with heavy ray tracing and NPC density. Dynamic MFG helped tame the stutter by keeping the generation mode at 3x or lower, which resulted in better latency than fixed 4x modes. However, the technology did not completely eliminate the janky feel associated with the heavy graphical load. CPU and GPU utilization figures showed that GPU usage remained high despite the performance lift provided by the AI.
Oblivion Remastered lacked native HDR support, but forced settings allowed for frame generation testing alongside Reflex. The PCL improved compared to baseline runs, yet the 1% low frame rate figures remained in a disappointing zone. This indicates that frame generation alone cannot fix underlying engine performance issues in older titles. Videos looked washed-out due to monitor configuration issues rather than rendering faults in the test environment.
The introduction of dynamic scaling marks a shift away from manual configuration toward automated optimization for gamers. While the technology effectively smooths visuals, the trade-off in input latency remains a critical factor for competitive play. Future updates may refine the AI models to reduce the perceived lag while maintaining high frame counts. Users must weigh the visual benefits against the responsiveness required for their specific genre preferences.
