Nvidia DLSS 1 and 2 antialiasing discussion *spawn*


That page just disappeared,... Still is found in google cache though:

https://webcache.googleusercontent....dia-reflex-support/+&cd=1&hl=en&ct=clnk&gl=us

From the cached copy
DLSS taps into the power of a deep learning neural network to boost frame rates and generate beautiful, sharp images. Reflex aligns CPU work to complete just in time for the GPU to start processing, minimizing latency and improving system responsiveness.

DLSS source code and NVIDIA Reflex is available now in Unreal Engine 5, and we will be releasing the DLSS plugin for UE5 in the coming weeks.
 
What does this actually mean? Is it another custom Nvidia build like with UE4, is DLSS native to UE5 or will devs need a 3rd party plugin to enable DLSS?

DLSS source code and NVIDIA Reflex is available now in Unreal Engine 5, and we will be releasing the DLSS plugin for UE5 in the coming weeks.
 
What does this actually mean? Is it another custom Nvidia build like with UE4, is DLSS native to UE5 or will devs need a 3rd party plugin to enable DLSS?

I read it as ue5 has the necessary plumbing implemented to make integrating dlss possible. Nvidia would release plugin enabling dlss in the coming weeks. Basically UE5 has to offer API that provides data and parameters needed by dlss plugin. Plugin approach is nice as there is no need to change engine/game to enable new feature(s)
 
I read it as ue5 has the necessary plumbing implemented to make integrating dlss possible. Nvidia would release plugin enabling dlss in the coming weeks. Basically UE5 has to offer API that provides data and parameters needed by dlss plugin. Plugin approach is nice as there is no need to change engine/game to enable new feature(s)

This makes sense. Hopefully from a developer perspective it’s just flipping a switch.
 
It was an easy way for PC gamers to dunk on console owners. It wasn't the opinion of everybody (especially not here). Also bad implementations deserve to be criticized. DLSS 1.0 was deservedly torn to shreds…
Not before the launch of FidelityFX CAS, which showed that DLSS (1.0) is worse than simple non-AI upsampling. Before that:
  • "DLSS does a great job" - tom'shardware
  • "results are impressive" - notebookcheck
  • "we're impressed with this technology" - eurogamer
  • "no loss in image quality seems almost too good to be true" - notebookcheck
  • "results look pretty impressive" - techradar
  • "impressive technological innovation" - overclock3d
  • "DLSS is quietly PC gaming's greatest graphical revolution" - vg247
  • etc.
No "torning to shreds" until the release of CAS/SR, just superlatives. No matter the image quality.
…so what you're saying isn't 100% true

The entire DLSS 1.0 era was like "DLSS is supergreat, AMD doesn't have built-in AI accelerators, AMD can't beat DLSS" and so on. Criticism was almost non existent. Most of the PC gamers and reviewers started hyping upsampling at the moment it arrived to the PC platform, irrespective of its quality. Upsampling, which was laughable to PC gamers until 2018, became the holy grail of PC gaming after the release of DLSS. Comparasions of quality weren't needed.

The change didn't come until the release of CAS. But instantly after the release of DLSS 2.0, a new idol was created. Again - without proper qualitative reviews. I remember just one review comparing quality of DLSS 2.0 to CAS, the rest hyped DLSS 2.0 without any qualitative / comparative testing. Again.

I don't know if FSR will be better, comparable or worse, or if RDNA3 will bring any better upsampling solution, but the acceptance of upsampling in PC space at 2018 wasn't related to some sort of superior quality (nobody cared, reviews lacked qualitative comparisions), but only because of the surrounding hype.

Anybody who disagree may feel free to post links to all the 2018 reviews comparing console upsampling / DLSS 1.0 upsampling quality. That would give some foundation to the statement, that the immediate change in the acceptance of upsampling in PC-space was quality-related and that all the PC gamers changed their minds by seeing the great visual improvement. Unfortunately I don't remember any single 2018 review comparing quality of DLSS to upsampling of consoles. It was more like Nvidia told to the media that DLSS is great because of
 
The change didn't come until the release of CAS. But instantly after the release of DLSS 2.0, a new idol was created. Again - without proper qualitative reviews. I remember just one review comparing quality of DLSS 2.0 to CAS, the rest hyped DLSS 2.0 without any qualitative / comparative testing.
If I recall there were comparative reviews which compared image quality, some which were used in this forum to support an argument one way or another.
https://forum.beyond3d.com/posts/2179713/
 
DLSS 1.0 was legitimately terrible and looked worse than basic monitor upscaling more often than not. That didn’t stop Nvidia fans from declaring it as an awesome, game changing technology. Hardware Unboxed was the only outlet i recall providing actual comparisons showing its uselessness.
 
DLSS 1.0 was legitimately terrible and looked worse than basic monitor upscaling more often than not.
It pretty much never looked worse than basic monitor upscaling when normalized for performance. HUB was the only source which made such nonsense claims.
What it did is introduce an obvious "ML hallucination" style into the upscaled image which may not have been to everyone's liking.
 
It pretty much never looked worse than basic monitor upscaling when normalized for performance. HUB was the only source which made such nonsense claims.
What it did is introduce an obvious "ML hallucination" style into the upscaled image which may not have been to everyone's liking.
Their claims were perfectly legitimate with video footage as evidence in both Metro and BFV. It had some benefits in FFXV due to a broken TAA implementation.
 
Their claims were perfectly legitimate with video footage as evidence in both Metro and BFV. It had some benefits in FFXV due to a broken TAA implementation.
I don't think that I have the energy to engage into this but their claims were about in-game resolution scaling being able to provide better quality at the same fps, not about "basic monitor upscaling" being able to do this as you've said.
There's a significant difference between the two in both quality and the availability.
Also their claims were shown in about two games AFAIR while there were more games with DLSS 1.0 and what they've said was hardly universal for all of them, especially since even in Metro it wasn't as clear cut as they made it sound.
For example they've used PP sharpening on downres example as an added benefit of it - as if you can't apply PP sharpening to DLSS 1.0 image.
 
Not before the launch of FidelityFX CAS, which showed that DLSS (1.0) is worse than simple non-AI upsampling. Before that:
  • "DLSS does a great job" - tom'shardware
  • "results are impressive" - notebookcheck
  • "we're impressed with this technology" - eurogamer
  • "no loss in image quality seems almost too good to be true" - notebookcheck
  • "results look pretty impressive" - techradar
  • "impressive technological innovation" - overclock3d
  • "DLSS is quietly PC gaming's greatest graphical revolution" - vg247
  • etc.
No "torning to shreds" until the release of CAS/SR, just superlatives. No matter the image quality.

A selection of gaming press articles that say differently doesn't prove your point. Other press, and posters on technical forums, were not impressed. DLSS 1.9 and 2.0 were Nvidia's response to the general shared disappoint in 1.0.


The entire DLSS 1.0 era was like "DLSS is supergreat, AMD doesn't have built-in AI accelerators, AMD can't beat DLSS" and so on. Criticism was almost non existent. Most of the PC gamers and reviewers started hyping upsampling at the moment it arrived to the PC platform, irrespective of its quality. Upsampling, which was laughable to PC gamers until 2018, became the holy grail of PC gaming after the release of DLSS. Comparasions of quality weren't needed.

Good grief :LOL: Our recollections clearly differ greatly.
 
3080 Ti looks like a boring product, unfortunately. It lacks the impacts of both 1080 Ti and 2080 Ti once made.
3070 Ti on the other hand might be more interesting.
 
Back
Top