The Borderless Gaming application has introduced an update that implements DLSS 5-style neural rendering and frame generation features, allowing them to run on hardware that NVIDIA has not officially cleared for the technology. This software-based implementation, developed by Andrew Sampson, utilizes a specific capture method that avoids modifying game files or interacting directly with NVIDIA’s proprietary software libraries.
NVIDIA officially launched DLSS 5 with 3D-Guided Neural Rendering on September 3, 2026. While the official version is restricted to the GeForce RTX 50-series (Blackwell) architecture and GeForce NOW Ultimate tiers, the Borderless Gaming update aims to democratize these features for users on older NVIDIA GPUs, as well as AMD and Intel hardware.
The Shift to Compositor-Level Rendering
Unlike traditional DLSS mods that require users to swap DLL files or inject code into a game’s directory, this implementation uses a “BGFX” mode. This method captures the game window at the compositor level—the part of the operating system responsible for drawing the final image on the screen. Because it operates outside the game’s internal engine, the tool can apply neural upscaling and frame generation to virtually any windowed application without risking anti-cheat triggers or compatibility crashes.
This approach is fundamentally different from native DLSS, which relies on motion vectors and depth buffers provided directly by the game engine. By processing the visual data at the compositor level, the tool acts as a post-processing layer. While this can introduce a slight increase in input latency compared to a native engine-level implementation, it offers a level of universal compatibility that official hardware-locked solutions do not provide.
Overcoming Hardware Limitations via Quantization
A primary reason NVIDIA cites for the RTX 50-series exclusivity of DLSS 5 is the requirement for specific architectural changes in the Blackwell tensor cores to handle 3D-Guided Neural Rendering. To bypass these hardware requirements, the Borderless Gaming implementation uses quantization.
Quantization involves reducing the precision of the neural network’s weights and activations—typically from FP32 or FP16 down to INT8. This reduction allows the complex math required for neural rendering to run efficiently on older hardware or general-purpose shaders that lack the specialized throughput of the newest GPUs. While quantization can occasionally lead to minor visual artifacts compared to full-precision models, it enables the “neural” aspect of the upscaling to function on a significantly broader range of hardware.
Market Impact and Availability
The release of this software-based alternative comes at a time when the high-end GPU market faces significant supply challenges. Prices for the flagship GeForce RTX 5090 have seen substantial inflation, reportedly driven by shortages in GDDR7 memory. By providing a pathway to DLSS 5 features on existing hardware, Borderless Gaming offers a performance-extending alternative for players unable or unwilling to pay premium prices for the newest generation of cards.
The custom frame generation feature included in the update was reportedly trained on tens of thousands of hours of gameplay data to ensure the motion interpolation remains fluid across different genres. As more developers move toward neural-heavy rendering pipelines, tools that decouple these features from specific hardware generations are likely to become increasingly relevant for maintaining the longevity of older gaming rigs.
