Meet the Two Modders Who Unlocked DLSS on Older GPUs
Dashdogy unlocked DLSS Multi Frame Generation to run on GPUs based on the Ada architecture after analyzing how Streamline and NGX work, eventually solving the black-screen issue caused by incompatible instructions.
Meanwhile, Uncle Burrito, an RTX Remix modder, analyzed the DLSS 5 Neural Rendering DLL and found that most of the code already supported Ada. He therefore ported it to run on GPUs from this generation. All of these modifications were made at the software level, eliminating the need to change hardware in order to try features originally restricted to newer GPUs.
From Casual Searching to Unlocking MFG on the RTX 40 Series
Dashdogy’s journey began when the RTX 5000 series launched in 2025. He had seen Nvidia demonstrate Multi Frame Generation, which was restricted to Blackwell, and decided to investigate whether it could run on the RTX 4000 series.
During his mid-semester break, he used an AWS instance and spent an entire week experimenting through trial and error. The main problem was that Blackwell and Ada Lovelace used incompatible instructions, causing the GPU to produce frames incorrectly until the screen turned black. He eventually found a way around the issue.
Features Locked Within Nvidia’s System
DLSS Multi Frame Generation (MFG) and DLSS 5 Neural Rendering are features in Nvidia’s AI software stack. Streamline acts as an instruction translator that allows games to call various technologies, while NGX is the backend system that handles the computations.
Originally, MFG was locked to GPUs based on the Blackwell architecture, or the RTX 5000 series. The first version of DLSS 5 Neural Rendering also did not allow Ada Lovelace GPUs to use it, although Nvidia later clarified that the RTX 4000 series was supported. This limitation therefore appeared to be more related to software restrictions and implementation than to hardware alone.
From the RTX 5000 Back to Older Architectures
MFG launched as a feature locked to Blackwell, or the RTX 5000 series. Ada Lovelace never officially received the feature, but Dashdogy unlocked it through a mod, and it has since been ported as far back as Turing, released in 2018.
| Factor | Blackwell | Ada Lovelace | Turing |
|---|---|---|---|
| Official MFG | Supported | Never released | Not specified |
| Mod status | No unlocking required | Unlocked | Ported |
| DLSS 5 NR | Supported | Nvidia confirmed support | Ported |
| Instruction limitations | Own instruction set | ISA differs from Blackwell | Not specified |
The key point is that Blackwell and Ada use different ISAs, which previously caused black-screen issues when Blackwell instructions were run on Ada.
When Games Demand More Frames Than the GPU Is Allowed to Produce
Dashdogy’s mod intercepts calls to Streamline and NGX in memory, allowing the RTX 40 Series to produce five frames instead of one. This helps avoid the black-screen issue that occurred when Blackwell instructions were run on Ada.
On the other hand, Uncle Burrito reverse-engineered the DLSS 5 Neural Rendering DLL and found that it could run on Ada, even though the first version did not provide complete support. There are now tools that make it easier to install both Neural Rendering and MFG in many games.
Behind the Scenes: Streamline and NGX
Streamline acts as an intermediary layer connecting games to multiple upscaling technologies. NGX is Nvidia’s software that processes AI features and sends the results back through Streamline to the game.
Dashdogy intercepts Streamline and NGX calls in memory, modifying the data that determines how many frames the graphics card can generate. This allows MFG to run on older cards through the software already being called by the game.
How Community Mods Differ from the Official System
Nvidia’s official system ties features such as MFG to the Blackwell architecture, preventing older cards from using them in the first place. Community mods instead reverse-engineer software restrictions and allow existing hardware to use these features.
| Factor | Nvidia approach | Community mod approach |
|---|---|---|
| Control point | Limited by architecture | Analysis of DLLs and calls |
| How it works | Enables features as defined | Adapts software for older hardware |
| Result | MFG tied to newer cards | Extends MFG and Neural Rendering to older models |
Uncle Burrito’s work also suggests that some of the code in DLSS 5 may already support Ada, but was incomplete in the first version.
What This Unlock Makes Possible—and What to Watch Out For
The mods extend the useful life of older GPUs and have enabled numerous forks and integrations, making it possible to install DLSS MFG and Neural Rendering in many games. They also support hardware going back to Turing, as well as Neural Rendering on AMD.
However, all of this remains unofficial modding. The ISA differences between Blackwell and Ada have caused black-screen issues in the past, and operation depends on intercepting Streamline and NGX instructions.
Pros
- +Extends the useful life of older GPUs
- +Enables a wide range of forks and integrations
- +Supports hardware going back to Turing
Cons
- −Unofficial mod
- −Different ISAs between Blackwell and Ada
- −History of black-screen issues
Costs That Do Not Appear on the Price Tag of the Graphics Card
The cost of modding does not end with the graphics card itself. It also includes the time required for reverse engineering and the cost of running an AWS instance for testing. Dashdogy spent quite some time working late at night before finally unlocking MFG.
DLSS 5 also required reverse-engineering a complex DLL and testing whether instructions from Blackwell could run on Ada. There was also a risk of encountering black screens caused by incompatible instructions.
After that, users still need community-maintained tools, forks, and integrations to install the mods in different games. Convenience therefore depends on community support rather than official support from Nvidia.
Who Should Try This Kind of Mod
These mods are suitable for owners of older GPUs who want to experiment with new features such as DLSS MFG or DLSS 5 NR, as well as people interested in game modding and software reverse engineering at the Streamline and NGX levels. Users should also understand that hardware limitations and incompatible instructions may cause black screens or prevent the features from working.
These are community mods, not new products that provide officially supported features after purchase. Installation therefore depends on community-maintained tools, forks, and integrations.
As the Line Between Hardware and Software Limitations Becomes Blurred
The work of Dashdogy and Uncle Burrito shows that features restricted to newer GPUs may not always be limited entirely by hardware. Some of the code may still support older hardware but remain disabled or unfinished.
The case of DLSS 5 NR makes this even clearer, as Nvidia later stated that the RTX 4000 series supported Neural Rendering. Reverse-engineering Streamline, NGX, and the DLLs therefore does more than unlock features—it also raises the question of how often future feature locks will be challenged through software analysis.