Preferences

> The image is rendered at a lower resolution, and then scaled up to native resolution, hopefully without noticeable artifacts or image quality degradation

That's what makes ray tracing not an attractive proposition at the moment. Once it will become usable without upscaling, then it will be interesting.


Why? DLSS Quality is virtually indistinguishable from native. No reason to not turn it on when available
I don't buy this "indistinguishable" claim. Especially in any general case. At most it would be not too annoying. It's always about paying with one quality for another.
Arguably native rendering wouldn't have antialiasing, because that would be downscaling, so DLSS would have both some disadvantages and some advantages. So the best case is better than "not too annoying", it has the potential to beat pure native rendering.
> Arguably native rendering wouldn't have antialiasing, because that would be downscaling

If you're unfamiliar with computer graphics it's unhelpful to try to nitpick and just end up being incorrect. Antialiasing is not downscaling, and less data is never going to beat more data for visual fidelity.

I said 'arguably' for a reason, not out of ignorance. Parts of the pipeline are still per-pixel, and parts are effectively rendering at a higher resolution.

> less data is never going to beat more data for visual fidelity

It has less data but better processing. That can look better sometimes.

Except it doesn't strictly have less data. It has multiple frames and other information to work with.

Frame generation techniques like TXAA and DLSS can provide superior quality to standard rasterization since they're taking advantage of extra information from previous frames, along with information that isn't fully exploited during traditional rasterization (like motion vectors and the depth buffer). DLSS goes a step further by exploiting the ML hardware on the GPU for increased throughput.

Of course if you set the internal resolution to 720p it's going to look bad, but there are cases where DLSS quite literally beats native when the internal resolution is close enough. And TXAA when implemented right is just 100% an improvement over non-TXAA in every case I've ever seen.

(I group TXAA and DLSS together because, generally speaking, they are the same sort of thing)

So it means that this temporal information can be used without upscaling to improve quality instead of mixing it with upscaling reducing it.
Bingo! God it's so frustrating when people mix the two up and end up thinking the whole world agrees with their conclusion that you can get a better image from less data, all else being equal. People still argue this all the time and it's absolutely maddening.
Agreed. It usually sounds like complete Koolaid.
Yeah, I think if you were to configure things to run DLSS with an internal resolution matching native, you would get enhanced quality in all cases. I don't know how feasible it is to do that though.

This item has no comments currently.

Keyboard Shortcuts

Story Lists

j
Next story
k
Previous story
Shift+j
Last story
Shift+k
First story
o Enter
Go to story URL
c
Go to comments
u
Go to author

Navigation

Shift+t
Go to top stories
Shift+n
Go to new stories
Shift+b
Go to best stories
Shift+a
Go to Ask HN
Shift+s
Go to Show HN

Miscellaneous

?
Show this modal