DLSS Neural Rendering Patcher

DLSS Neural Rendering Patcher is a command-line tool that prepares your copy of nvngx_dlssnr.dll for NVIDIA GeForce RTX 20, 30, and 40 Series GPUs.

The patcher checks its work before writing the patched file. It does not include, download, or distribute NVIDIA DLLs or model files.

Safety warning: Patching changes the DLL and invalidates its NVIDIA digital signature. Anti-cheat or file-integrity systems may flag the modified file. Do not use it with online, competitive, or anti-cheat-protected software.

Requirements

• Python 3.10 or newer
• CUDA Toolkit 13.3
  • Windows: If the CUDA installer reports that Visual Studio is missing, you can ignore the message and continue the installation. Visual Studio is not required to run this patcher.
  • The CUDA tools ptxas, fatbinary, and cuobjdump must be available on your system's PATH. If the patcher cannot find them, use --cuda-bin to point it to the CUDA Toolkit's bin folder.
• Your own original copy of nvngx_dlssnr.dll

Basic use

1. Download this repository as a ZIP file and extract it.

2. Find your original nvngx_dlssnr.dll and save a separate copy somewhere safe.

3. Open Command Prompt, PowerShell, or a terminal in the patcher's folder.

4. Run:

   python dlssnr_patcher.py "C:\path\to\nvngx_dlssnr.dll"
   

Replace the example path with the full path to your DLL. Keep the quotation marks, especially if the path contains spaces.

With no GPU options, the patcher includes every supported generation. If you are unsure which option to use, run the basic command above.

Choosing GPU generations

To select specific GPU generations instead of all four, add one or more of these options:

GeForce family	GPU generation	Option
RTX 20 Series	Turing	-t, --turing
RTX 30 Series	Ampere	-A, --ampere
RTX 40 Series	Ada	-a, --ada
RTX 50 Series	Blackwell	-b, --blackwell

Options can be combined. For example, this command selects Turing and Ampere:

python dlssnr_patcher.py --turing --ampere "C:\path\to\nvngx_dlssnr.dll"

Output and backups

By default, the patcher renames the original file to nvngx_dlssnr.dll.bak, then saves the patched file as nvngx_dlssnr.dll. It will not overwrite an existing backup.

Use --output patched.dll to keep the input file unchanged. Use --dry-run to check the DLL and compile the changes without writing an output file.

Run python dlssnr_patcher.py --help to see every option.

Contributing

Bug reports and focused pull requests are welcome. Read CONTRIBUTING.md before getting started.

License

Licensed under the GNU General Public License v2.0.