Modders have managed to get an experimental version of NVIDIA’s DLSS 5 Neural Rendering technology running in Control, only hours after a previously unseen DLSS-related DLL was discovered inside the PC early-access build of NBA 2K27.
The work comes from members of the RenoDX modding community, who used a newly discovered file named nvngx_dlssnr.dll to produce an unofficial implementation inside Remedy Entertainment’s 2019 action game. VideoCardz documented the experiment on August 27, while footage shared by the community shows Neural Rendering being applied to Jesse Faden’s character model.
The DLL was first found in NBA 2K27 and is identified by Windows as NVIDIA DLSSNR version 310.8.0.0. It is approximately 158 MB in size and carries an August 26 modification date. The “DLSSNR” naming aligns with NVIDIA’s own terminology for DLSS Neural Rendering, although NVIDIA has not publicly released this particular runtime as a standalone consumer feature.
In the current Control test, RenoDX modders have reportedly exposed seven visual styles. The effect is not being applied across the entire game, however. The demonstration is currently limited primarily to character models, making this an early technical experiment rather than a complete DLSS 5 integration.
There are several other limitations. HDR is not functioning correctly in the current implementation, with testing largely focused on SDR while the modders investigate how the Neural Rendering model expects its image data to be presented.
Performance is also far from representative of a finished NVIDIA implementation. RenoDX contributor speedlemur reported that one 4K test on a GeForce RTX 5070 Ti resulted in an almost 50% performance reduction. The current Neural Rendering pass is reportedly being inserted late in the rendering pipeline, after upscaling and potentially after tone mapping, which may contribute to the heavy overhead.
That figure should therefore not be treated as an indication of how DLSS 5 will perform when officially integrated into supported games. NVIDIA’s own implementation will be built into game engines with developer-controlled access to the information required by the model, rather than being injected experimentally through a community mod.
One notable detail from the RenoDX testing is that the Neural Rendering runtime appears able to operate using much of the data already available to existing DLSS features. That could make older titles an interesting target for experimentation if modders can continue adapting the technology without direct developer support.
The community has already begun experimenting beyond Control, with A Hat in Time also referenced as another title being tested with the new runtime. That game originally uses DirectX 9, making the experiment especially notable because it was never designed around modern NVIDIA neural rendering features.
NVIDIA officially announced DLSS 5 in March 2026. The company describes it as a real-time neural rendering system capable of modifying lighting, materials and other visual characteristics using an AI model, while still giving developers control over where and how the effect is applied.
Unlike familiar DLSS features focused primarily on reconstruction or frame generation, DLSS 5 is intended to alter the rendered appearance of scenes and individual objects. NVIDIA has said the technology is scheduled to launch this fall for GeForce RTX 50 Series GPUs.
The discovery of the DLSSNR runtime in NBA 2K27 is therefore significant because it gives modders access to a working component of the technology ahead of its official rollout. It does not prove that NBA 2K27 itself is using DLSS 5 in gameplay, nor does the Control experiment represent NVIDIA’s final image quality or performance.
For now, the RenoDX test is best viewed as an early look at what enthusiasts can extract from the leaked runtime. The fact that Neural Rendering has already been made to function inside an unsupported game suggests DLSS 5 could become another major target for graphics modding once NVIDIA releases the technology more broadly.

