RTX 20 Series DLSS Frame Generation Mod Shows Mixed Results
A community project called DLSSG for SM75 is attempting to bring NVIDIA’s native DLSS Frame Generation to RTX 20-series graphics cards, despite the feature never being officially supported on Turing hardware.
The project has already reached a working proof-of-concept stage. Testing on an RTX 2060 Max-Q with the DirectX 12 version of The Witcher 3: Wild Hunt Next-Gen update demonstrates that the in-game Frame Generation option can be enabled and produce measurable frame-rate improvements.
However, the project is still experimental. While its native calling mechanism is technically significant, early testing shows image-quality artifacts, scaling limitations, game compatibility issues, and performance that may not consistently outperform alternative solutions based on AMD’s FSR Frame Generation.
🚀 Native DLSS Frame Generation on Turing #
The key distinction between DLSSG for SM75 and earlier community Frame Generation solutions is how the mod handles NVIDIA’s DLSS runtime.
Previous mods targeting RTX 20-series and RTX 30-series GPUs commonly redirected Frame Generation requests to AMD’s FSR 3 implementation. DLSSG for SM75 instead attempts to execute NVIDIA’s own nvngx_dlssg 310.1 runtime directly.
This makes the project particularly interesting from a technical perspective: it is not simply replacing NVIDIA’s Frame Generation pipeline with another vendor’s implementation, but attempting to make the official NVIDIA runtime execute on an architecture outside its intended support range.
Turing and SM75 #
NVIDIA’s RTX 20-series GPUs are based on the Turing architecture and expose the SM75 compute capability.
The project builds on earlier community work that enabled DLSS Frame Generation experiments on RTX 30-series hardware. The basic approach involves adapting GPU kernels for SM75 and altering architecture-detection behavior so the runtime can proceed through initialization checks that would otherwise reject a Turing GPU.
In effect, the compatibility layer makes the older GPU appear sufficiently compatible with the runtime for the native Frame Generation path to initialize.
🧩 Official NVIDIA Support Remains Unchanged #
The existence of a working community implementation does not change NVIDIA’s official hardware support requirements.
Standard DLSS Frame Generation remains officially associated with the RTX 40 series and newer, while DLSS Multi Frame Generation (MFG) is restricted to the RTX 50 series.
The distinction between functional and supported is important here. A community modification can demonstrate that portions of a technology can execute on older hardware, but that does not mean NVIDIA validates the implementation, guarantees compatibility, or provides driver and game-level support for it.
For developers and enthusiasts, the project is therefore better understood as an experimental compatibility breakthrough rather than an unofficial extension of NVIDIA’s product support matrix.
⚠️ Early Performance and Image-Quality Limitations #
The current implementation remains far from production-ready.
Initial testing suggests that 2x Frame Generation is comparatively stable, while higher generation factors expose significant limitations. At 3x Frame Generation, users can encounter visible ghosting and other image-quality artifacts. 4x Frame Generation currently fails to operate.
These limitations indicate that successfully invoking the DLSS Frame Generation runtime is only one part of the problem. Proper frame-generation behavior also depends on the GPU’s execution characteristics, optical-flow processing, frame pacing, game integration, and the quality of the generated intermediate frames.
Native Does Not Automatically Mean Faster #
The most notable caveat is that native execution of NVIDIA’s DLSS Frame Generation runtime does not automatically guarantee superior performance.
Early results indicate that the Turing implementation can still fall behind FSR-based Frame Generation solutions in some scenarios. The additional compatibility work required to execute the newer runtime on an older architecture may introduce overhead, while the RTX 20-series hardware lacks the architectural capabilities and performance headroom of newer generations.
Consequently, the technical achievement of running the native NVIDIA pipeline should be separated from the practical question of whether it provides the best gaming experience.
🎮 Game Compatibility Remains Limited #
Game compatibility is another unresolved issue.
Community testing has already identified cases where the mod does not work correctly, and the current number of tested titles is too small to establish reliable compatibility across modern DX12 games.
DLSS Frame Generation is not purely a standalone GPU feature. Games must expose the appropriate rendering and motion-vector data, integrate the relevant NVIDIA libraries, and interact correctly with the frame-generation pipeline. Differences in engine implementations can therefore produce substantially different results from one title to another.
As testing expands, compatibility across engines and individual game releases will be one of the most important indicators of whether DLSSG for SM75 can evolve beyond a proof of concept.
🔬 A Significant Turing Modding Experiment #
DLSSG for SM75 is technically noteworthy because it challenges a long-standing hardware boundary imposed on NVIDIA’s Frame Generation stack.
Getting nvngx_dlssg 310.1 to initialize on an SM75 GPU demonstrates that at least part of the software stack can be adapted to hardware NVIDIA does not officially target. The experiment also highlights the role of runtime architecture checks and GPU kernel compatibility in determining whether newer graphics features can be made to operate on older hardware.
Nevertheless, the current results do not suggest that RTX 20-series owners are about to receive a fully mature replacement for officially supported DLSS Frame Generation.
For now, DLSSG for SM75 remains an experimental native DLSS Frame Generation implementation. Its 2x mode shows promise, but 3x and 4x generation remain problematic, game compatibility is still uncertain, and performance can trail FSR-based alternatives.
The project is therefore most valuable as a demonstration of what can be achieved by modifying NVIDIA’s software stack—not yet as a dependable general-purpose Frame Generation solution for Turing GPUs.