MFG Unlock

A ReShade addon that enables DLSS multi-frame generation (3x / 4x and above) on GeForce RTX 40-series cards, which NVIDIA ships gated to RTX 50-series only — and corrects the frame interpolation so the extra frames carry new motion instead of repeats.

Nothing on disk is modified. Every patch is applied to the mapped image at runtime and reverted when the addon unloads. No NVIDIA files are redistributed here.

---

Status

Game	Reaches	Notes
Cyberpunk 2077	6x	Has its own 2x/3x/4x selector; the addon can force beyond it
Deep Rock Galactic	6x	FG is on/off only, so the addon drives the count entirely. Needs a modern nvngx_dlssg.dll (see below)
Grand Theft Auto V Enhanced	4x	Genuine ceiling — its bundled sl.dlss_g 2.9.1.0 clamps to 3 generated frames

Other titles are likely to work; these three are what has actually been tested.

Requirements

• GeForce RTX 40-series. See Why not 30-series? below.
• ReShade with addon support (this is an .addon64, not an effect).
• A game shipping DLSS frame generation via Streamline, with a reasonably modern nvngx_dlssg.dll (310.x). Games still on the DLSS 3 snippet (3.5.x) contain no multi-frame code at all and need a newer one dropped in beside the executable.

Install

1. Install ReShade into the game, with addon support enabled.
2. Drop renodx-mfgunlock.addon64 next to the game executable.
3. Launch, open the ReShade overlay, and find MFG Unlock in the Add-ons tab.

Settings

Written to your ReShade.ini under [RenoDX.MFGUnlock]:

Key	Default	Meaning
Enabled	1	Master switch for the whole addon
MaxCount	4	The DLSSG.MultiFrameCountMax value reported to the runtime
ForceFlipMeteringOff	1	Required for 3x+ on Ada. Leave on
TemporalFix	1	The interpolation correction. Leave on
ForceMultiplier	0	0 respects the game's own choice; 2–6 forces that multiplier
RaiseFrameCeiling	0	Raises the Streamline plugin's internal clamp. Off by default because it breaks some games
ForceOTAPlugins	0	Asks Streamline to load the driver's OTA plugin set. Off by default; see notes

If a game has its own multiplier selector, leave ForceMultiplier at 0 and use the game's setting.

How it works

Three gates decide whether multi-frame generation is available, and the addon opens the two that matter:

1. nvngx_dlssg.dll exports NVSDK_NGX_GetGPUArchitecture as a hardcoded minimum architecture — mov eax, 0x190 (Ada). A 40-series card already clears this, so it is left alone.
2. DLSSGInstanceManager::PopulateParameters compares the NVAPI arch id against 0x1b0 (Blackwell) to decide whether to advertise a max frame count of 5 or 1.
3. A second compare against the same constant feeds a runtime capability flag that drives generation itself.

Patching (2) without (3) makes the options appear and then render black. The addon rewrites both compares, in both encodings, in memory only — NGX verifies the snippet's Authenticode signature at load time, so the same bytes changed on disk make frame generation disappear entirely.

Unlocking the count alone is not enough. The interpolation kernel blends with a compiled-in 0.5, so every generated frame lands at the temporal midpoint: 4x produces three identical half-way frames, the counter doubles and the motion does not get smoother. The addon decompresses the kernel's PTX, rewrites the blend weight to come from the kernel's own temporal parameter, and re-emits the fatbin so the driver JITs the corrected version.

Blackwell paces multi-frame output with hardware flip metering Ada does not have; left enabled it freezes the presented image. Streamline already ships a software fallback, so the addon forces the plugin down it.

Each source file documents its own area in detail — start with the header comment in addon.cpp.

Why not 30-series?

Not because of the gates — those are just constants. Because the DLSS 4 snippet ships no Ampere machine code. Its 70 fatbins carry PTX sm_89 ×70, PTX sm_120 ×31 and cubin sm_89 ×31, and nothing for sm_80/sm_86. PTX is forward-compatible only, so sm_89 PTX cannot be JIT-compiled down to sm_86; the module load fails outright.

Retargeting is theoretically open — the kernels use only mma.sync m16n8k16/m16n8k8 FP16 and ldmatrix, with zero instructions newer than sm_86 (no FP8, no wgmma, no TMA), and the old hardware optical-flow dependency is gone in DLSS 4. But frame generation costs roughly a fixed amount per generated frame, and Ampere has far less FP16 tensor throughput per SM, so the generation pass would likely cost more than the frame it saves. It was investigated and deliberately dropped.

Building

The addon is built as part of a RenoDX tree, which supplies ReShade, ImGui, Detours, and the NGX/Streamline headers.

git clone --recursive https://github.com/clshortfuse/renodx
cp -r src/addons/mfgunlock <renodx>/src/addons/
cd <renodx>
cmake --preset vs-x64
cmake --build build.vs --config Release --target mfgunlock

The build globs src/**/**/addon.cpp, so no CMake changes are needed. The output is build.vs/Release/renodx-mfgunlock.addon64.

Prebuilt binaries are attached to Releases.

Credits

• The midpoint diagnosis, the injected PTX, and the fatbin truncation trick come from dashdogy/RTX40MFG-Unlock. That project ships no licence, so nothing here is copied from it — midpoint.hpp is an independent implementation of the same idea, verified by reproducing its published output digest byte-for-byte.
• Built on RenoDX by clshortfuse, and ReShade by crosire.

Disclaimer

Not affiliated with or endorsed by NVIDIA. This modifies process memory of a running game; use it on your own hardware at your own risk, and expect anti-cheat in multiplayer titles to object. Results on hardware NVIDIA did not ship this feature for are to be judged by eye.

Licence

MIT — see LICENSE.