DLSS Not Working on Linux. NVIDIA Drivers installed

[TheRealBlarp]

I'm running ubuntu linux with nvidia drivers installed and an rtx 3050 graphics card. Reflex and DLSS aren't working. Should I just install windows?

[Stahlmann]

I don't have any Linux experience, but I've heard that Nvidia can be a real pita on Linux. If your PC is mostly for gaming, then I'd recommend to just install Windows.

[demonix00]

I wouldn't even bother installing windows for better DLSS support since it's barely going to function on a 3050.

[Mark Kaine]

6 hours ago, demonix00 said:

I wouldn't even bother installing windows for better DLSS support since it's barely going to function on a 3050.

i think the opposite actually, you'd want it on a low range card as every little bit of performance can help.

 

dlss doesn't take resources it *gives* you more headroom as it lowers quality  = hence gives you more performance. 

 

 

[Stahlmann]

19 hours ago, demonix00 said:

I wouldn't even bother installing windows for better DLSS support since it's barely going to function on a 3050.

DLSS is just as useful on low-end cards as it is on high-end cards. I think you're confusing DLSS with ray tracing, and I agree. Ray tracing performance will be so bad on a 3050 that I probably wouldn't even call it a feature.

[xAcid9]

For DLSS, try use DXVK_ENABLE_NVAPI=1 %command% parameter. 

Reflex is not supported iirc. You have LatencyFlex though. 

[demonix00]

On 5/8/2024 at 11:57 AM, Stahlmann said:

DLSS is just as useful on low-end cards as it is on high-end cards. I think you're confusing DLSS with ray tracing, and I agree. Ray tracing performance will be so bad on a 3050 that I probably wouldn't even call it a feature.

Nope, I'm talking about DLSS being utter trash on low end RTX cards because of the following...

 

 

 

Now you could call those results a margin of error, but then shouldn't the DLSS score while using an RTX graphics card by much higher then the FSR score.

[Stahlmann]

10 hours ago, demonix00 said:

Nope, I'm talking about DLSS being utter trash on low end RTX cards because of the following...

 

Now you could call those results a margin of error, but then shouldn't the DLSS score while using an RTX graphics card by much higher then the FSR score.

No, DLSS is not better than FSR in all cases. But even if it performs slightly worse on the same quality preset, the image quality and especially the image stability (meaning in terms of flickering, shimmering, aliasing, etc.) is still superior to FSR in almost all situations, especially at lower target resolutions like 1080p and 1440p. The difference isn't as big at 4K, but it's still there.

 

I don't have any personal experience with Final Fantasy 14, so I don't know how a difference of 756 points translates into FPS. The difference is about 7%, so if that translates into 7% better FPS, I wouldn't call DLSS "utter trash" because of such a small difference.

[demonix00]

On 5/10/2024 at 8:41 AM, Stahlmann said:

No, DLSS is not better than FSR in all cases. But even if it performs slightly worse on the same quality preset, the image quality and especially the image stability (meaning in terms of flickering, shimmering, aliasing, etc.) is still superior to FSR in almost all situations, especially at lower target resolutions like 1080p and 1440p. The difference isn't as big at 4K, but it's still there.

 

I don't have any personal experience with Final Fantasy 14, so I don't know how a difference of 756 points translates into FPS. The difference is about 7%, so if that translates into 7% better FPS, I wouldn't call DLSS "utter trash" because of such a small difference.

At least I saved the results from those two benchmark runs (and this isn't like the benchmarks that most games have where it just does a flyby of a scene and gives a result, it goes through five different scenes with one of the scenes being of active combat with a large number of characters involved along with all the combat effects)

 

DLSS score 10326

Average FPS: 76.97476

Minimum FPS: 32

 

FSR score 11082

Average FPS: 83.40252

Minimum FPS: 34

 

Also the claims that nvidia make on their own site that DLSS can almost double your FPS on any RTX card (if you look at the section for DLSS, it only states the resolution and game setting and doesn't mention the specifications of the PC used at all) doesn't hold water on lower end cards since someone else did mention they got higher results using DLSS over FSR because they had a higher end nvidia GPU, which is why I told the OP that installing windows just for better DLSS support would be pointless because the RTX3050 they have doesn't have the grunt needed to make full use of DLSS meaning that it's as I stated utter trash on lower end cards.

[Stahlmann]

13 hours ago, demonix00 said:

At least I saved the results from those two benchmark runs (and this isn't like the benchmarks that most games have where it just does a flyby of a scene and gives a result, it goes through five different scenes with one of the scenes being of active combat with a large number of characters involved along with all the combat effects)

 

DLSS score 10326

Average FPS: 76.97476

Minimum FPS: 32

 

FSR score 11082

Average FPS: 83.40252

Minimum FPS: 34

So the difference was as I expected. 76 vs 83 FPS isn't a big deal and you still have the choice between slightly better FPS or slightly better image quality.

 

13 hours ago, demonix00 said:

Also the claims that nvidia make on their own site that DLSS can almost double your FPS on any RTX card (if you look at the section for DLSS, it only states the resolution and game setting and doesn't mention the specifications of the PC used at all) doesn't hold water on lower end cards since someone else did mention they got higher results using DLSS over FSR because they had a higher end nvidia GPU, which is why I told the OP that installing windows just for better DLSS support would be pointless because the RTX3050 they have doesn't have the grunt needed to make full use of DLSS meaning that it's as I stated utter trash on lower end cards.

These claims pretty much only hold water when you use ultra performance presets, but at that point image quality also severely suffers. But that's also the case for FSR. The main reason I can think of why DLSS doesn't improve as much as FSR on lower-end rigs is probably because of the CPU-sided sheduler that makes it so Nvidia cards need slightly more CPU performance than AMD ones, and lower end rigs are CPU limited faster.