A $5 Starfield mod to enable DLSS Frame Generation has been cracked to remove modder's DRM

[DukenukemX]

GPUOpen’s team leaders have said publicly and on camera that as long as DLSS is closed source they will do everything they can to ensure that FSR and DLSS do not coexist in the same development package. So FSR 2 and up will be blocked from Streamline and DLSS will be blocked from GPUOpen.

It would make more sense for Nvidia to open source DLSS than for AMD to just bless them to include FSR.

So ultimately while the actual functionality of DLSS is closed, AMD and the GPUOpen project have a very real and easy method for blocking it's integration.

Why is it important if AMD or Nvidia are responsible for distributing each others upscaling technology? I'd rather neither of them do because it'll always result in one fucking over the other in some way.

 

[Lakados]

It would make more sense for Nvidia to open source DLSS than for AMD to just bless them to include FSR.

Why is it important if AMD or Nvidia are responsible for distributing each others upscaling technology? I'd rather neither of them do because it'll always result in one fucking over the other in some way.

It would make sense but Nvidia puts their secret sauce right there in the code, they hard value like nobody’s business and it would 100% give away their secret sauce on everything. Nvidia needs a full rewrite on their drivers to do this and a shift in how they build their cards.
Nvidia firmware is quite lite, because it’s all in the drivers, AMD at least puts that part in a single binary firmware blob for their open source drivers. NVidia could do the same but it’s going to take more work than they can crank out in a weekend.

They have sort of open sourced their drivers, in that the calls are now open with closed blobs littered all over the place.

Nvidia is very likely going to do it, at least with their creator and Linux drivers, we’ll see if they do for the Windows gaming ones.

DLSS should go open, but as it’s a CUDA front end it wouldn’t matter even if it did. Nobody else can run it so it’s a moot point.

FSR and XeSS need to do better and CUDA needs to get supplanted.

Special things like tensor cores or accelerators need to be standardized to some degree just as the rest of the pipeline is. This is a DirectX Next problem

Edit:
I am hopeful FSR 3 can bring things to a point where it works well enough to be an acceptable alternative to DLSS. A full HLSL implementation has the possibility of doing anything CUDA can (in regards to graphics) while being something Microsoft and Kronos could feasibly roll into the official dev libraries as a supported feature and it runs on the existing architectures. IF that happens (Big ass IF) then AMD, Nvidia, and Intel could work to optimize that on their hardware at a driver level and assign it to whatever onboard processes they feel makes sense, it's the closest thing to a peaceful resolution I see happening, everything else has developers and us along with them getting dicked around for marketing clout.