Tested driver: 616.92 DCH (GeForce, desktop, RTX 3070, Windows).
Applied the byte-verification check from win/win10_x64/616.92/nvencodeapi64.1337 and nvencodeapi.1337 (verifying each offset:old->new entry against the live DLL before writing anything, rather than patching blind).
Result: 0/7 offsets match in nvencodeapi64.dll, and the x86 patch for nvencodeapi.dll also fails to match.
To rule out local corruption or leftover files, I verified:
- Clean reinstall of the official installer (616.92-desktop-win10-win11-64bit-international-nsd-dch-whql.exe), no GeForce Experience, custom install.
- SHA256 of the live C:\Windows\System32\nvencodeapi64.dll: 06a9f8e8dfcc64c71f00735dd0063e8a69ea4fc1f2ca7b1469ca47bdaf12f00e
- SHA256 of Display.Driver\nvencodeapi64.dll extracted directly from that same official installer package: identical (06a9f8e8dfcc64c71f00735dd0063e8a69ea4fc1f2ca7b1469ca47bdaf12f00e).
So the installed file is byte-for-byte identical to what NVIDIA ships in the official installer - this isn't a corrupted or modified local file. The published patch offsets for 616.92 in this repo just don't correspond to this build.
For context, I hit the exact same 0/7 result on driver 596.36 first (added in #1064 / #1066) before trying 616.92, so this may not be isolated to a single version.
Happy to provide the extracted DLL or more details if it helps find the correct offsets.
Tested driver: 616.92 DCH (GeForce, desktop, RTX 3070, Windows).
Applied the byte-verification check from win/win10_x64/616.92/nvencodeapi64.1337 and nvencodeapi.1337 (verifying each offset:old->new entry against the live DLL before writing anything, rather than patching blind).
Result: 0/7 offsets match in nvencodeapi64.dll, and the x86 patch for nvencodeapi.dll also fails to match.
To rule out local corruption or leftover files, I verified:
So the installed file is byte-for-byte identical to what NVIDIA ships in the official installer - this isn't a corrupted or modified local file. The published patch offsets for 616.92 in this repo just don't correspond to this build.
For context, I hit the exact same 0/7 result on driver 596.36 first (added in #1064 / #1066) before trying 616.92, so this may not be isolated to a single version.
Happy to provide the extracted DLL or more details if it helps find the correct offsets.