What is the difference Lossless Scaling and dlss 4

[ISAM]

Title

[Space Detective]

DLSS generally requires explicit in-game support, which you'll probably only see in big modern mainstream titles.
LS works with basically anything, from games to emulators to browsers to video players.

DLSS needs Nvidia GPUs, and specific ones at that.
LS is GPU-agnostic, and only really cares about how performant your GPU is.

DLSS maxes out at 4x frame gen, from what I hear.
LS can do all the way up to 20x, if your GPU can handle it.

DLSS can make use of game data to improve the visual results of frame generation (i.e. reduce artifacting).
LS can only work with the final frame when generating in-between frames, so it can be more prone to artifacts.

DLSS, basically just works without any need for you to put in effort as long as you understand not to overload your GPU.
LS is a bit more involved and complex to get the most effective use out of. Not massively so, but there is a learning curve.

DLSS can make use of tensor cores to accelerate itself.
LS cannot, at all, even if you have one of the GPUs that have them.

I think that's the gist of it, as far as frame gen stuff is concerned.

[Ramstar]

This works for increasing fps.

[Gizzmoe]

Originally posted by Space Detective:

DLSS can make use of tensor cores to accelerate itself. LS cannot, at all, even if you have one of the GPUs that have them.

Just to avoid confusion for some, it 's not that LS *cannot* make use of tensor cores, tensor core support simply hasn't been implemented into LS. We'll see if that's going to change in future versions.

[Gizzmoe]

Originally posted by Space Detective:

DLSS maxes out at 4x frame gen, from what I hear.

x3/x4 framegen only for 5xxx series, x2 for 4xxx.

[Space Detective]

Originally posted by Gizzmoe:

Just to avoid confusion for some, it 's not that LS *cannot* make use of tensor cores, tensor core support simply hasn't been implemented into LS. We'll see if that's going to change in future versions.

Yeah, I meant to speak for its current state rather than as an absolute.

x3/x4 framegen only for 5xxx series, x2 for 4xxx.

Ah yeah, right.

[Steel]

Originally posted by Gizzmoe:

Originally posted by Space Detective:

DLSS can make use of tensor cores to accelerate itself. LS cannot, at all, even if you have one of the GPUs that have them.

Just to avoid confusion for some, it 's not that LS *cannot* make use of tensor cores, tensor core support simply hasn't been implemented into LS. We'll see if that's going to change in future versions.

That would be a killer update. But I wonder how feasible it is? Nvidia in the past hasn't been famous for open source documentation and code, but I'm out of the loop there.