I can't get my head around W1zzard's huge praise for DLSS 3.5 Ray Reconstruction.
"Better than native" with lots of blurb which seems to have been written by Nvidia's PR department.
The image comparisons tell a different story IMO. Especially the last image:
NVIDIA DLSS 3.5 launches today, supporting GeForce 20 and newer. The new algorithm improves the looks ray traced lighting, reflections, shadows and more. In our DLSS 3.5 Ray Reconstruction review we test image quality, but also discovered that VRAM usage actually goes down and performance goes up.
www.techpowerup.com
Better than native: because who wants to see the marble texture on the floor anyhow? Nice and blurred and soft is better?
Better than native: the reflections of those lanterns needs to do soft and blurry because that's all RR can do?
Lighting is a bit better, but that's it. Maybe all those years of testing GPUs has made W1zzard's unable to see textures?
Depends on the context entirely.
It's noted by DF and others, as well as nvidia although their wording could have been better than "Better than native", it's about the ray tracing de-noising. Since RR uses tensor cores, and as such DLSS has to be enabled for it to work, this results in better than native ray tracing as you no longer have the "walking pixels" noise in many areas of a ray traced scene like in shadows or on car paintwork under say shadow or on fences. So yes in that regard it /is/ better than native, until the day comes that ray reconstruction is possible at native res which bypasses the AI accelerator cores, then
clean ray tracing will always be better than ray tracing at native res. Also on top of that DLSS3.5 removes the light response latency almost entirely which is another big positive gain as lighting reacts almost instantly now.
It can be fairly argued well how important is clean RT de-noising + the removal of the long light response latency vs say slightly less detailed reflections on a floor? Clean de-noising is a significant improvement for the overall visual experience in games that have ray tracing (currently only path tracing until the AI models are trained on normal RT too).
You also have to show a GIF or video side by side to capture what clean ray tracing is like vs native ray tracing, a still image is not suitable for that sort of comparison.
There will no doubt be an argument from those against the use of RT anyway since screen space looks good enough in this game, and in many situations I would agree, if you can't run RT/PT with acceptable fps, then there are only two options, upgrade the GPU or use lower settings.
The comparisons with RT vs PT vs PT+RR vs all RT off is quite significant and has been done to death already. What you will find is typically cherry picked examples where the difference is much smaller, and those are due to the fact that in those scenes at that in-game time of day, actual light would look similar and screen space approximations are as such, accurate enough to be convincing against RT/PT.
But in all other cases where actual light and shadow play a real part, there is only one real winner.
How important it is will be down to personal preference (and hardware obviously). No point running PT if the performance is crap, for example.
Check it out... Click the first screenshot, then use the arrow keys to cycle between them to best see the diff between PT+RR vs PT and no RR.
So yeah, Ray reconstruction = higher fps, higher detail, higher quality detailed lighting.