posted in Technology

GolfFoxtrotLima@GolfFoxtrotLima@sh.itjust.works

DLSS 5 comes with a massive 50-60% performance hit, so I hope you like lower frame rates with your dose of AI image enhancement

cross-posted from: lemmy.world/post/51453544

The very bad news is, Nvidia says turning DLSS 5 on will currently reduce performance by 50-60%. And that’s across the board—no matter the game.

www.pcgamer.com/hardware/graphics-cards/dlss-5-comes-with-a-massive-50-60-percent-performance-hit-so-i-hope-you-like-lower-frame-rates-with-your-dose-of-ai-image-enhancement/
givesomefucks@givesomefucks@lemmy.world

Replying to @⁨GolfFoxtrotLima@sh.itjust.works⁩

Genuinely asking:

Is this the first time DLSS hurts fps instead of improving it?

Like, I thought that was the entire point of it. It allowed you to turn up settings but still have useable frame rates.

I don’t know who would pay for a card to support this, just to turn around and cut the fps in half.

It seems like their target demo is people that pay for 5090s but for some reason still put on all the frame gen and scaling stuff to make their game look worse. This chops the frame in half, when they could just turn off frame gen and scaling and likely have better and steadier frames.

en
urushitan 漆たん@urushitan@kakera.kintsugi.moe

Replying to @⁨Dudewitbow@lemmy.zip⁩

DLAA is the anti aliasing only. DLSS was always ai upscaling with anti aliasing as a first pass, literally Deep Learning Super Sampling.

DLSS 1

The first iteration of DLSS is a predominantly spatial image upscaler with two stages, both relying on convolutional auto-encoder neural networks.[31] The first step is an image enhancement network which uses the current frame and motion vectors to perform edge enhancement, and spatial anti-aliasing. The second stage is an image upscaling step which uses the single raw, low-resolution frame to upscale the image to the desired output resolution. Using just a single frame for upscaling means the neural network itself must generate a large amount of new information to produce the high-resolution output, which can result in slight hallucinations such as leaves that differ in style to the source content.[13]

en.wikipedia.org/…/Deep_Learning_Super_Sampling