NVIDIA DLSS 2.0: Using AI to achieve max graphics fidelity
What is DLSS?
“Deep Learning Super Sampling (DLSS), NVIDIA set out to redefine real-time rendering through AI-based super resolution – rendering fewer pixels and then using AI to construct sharp, higher resolution images.” – Nvidia
DLSS 2.0 offers several key enhancements over the original version:
- Superior Image Quality – DLSS 2.0 offers image quality comparable to native resolution while rendering only one quarter to one half of the pixels. It employs new temporal feedback techniques for sharper image details and improved stability from frame to frame.
- Great Scaling Across All GeForce RTX GPUs and Resolutions – A new AI network more efficiently uses Tensor Cores to execute 2X faster than the original. This improves frame rates and eliminates previous limitations on which GPUs, settings, and resolutions could be enabled.
- One Network For All Games – The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalized network that works across games. This means faster game integrations, and ultimately more DLSS games.
- Customizable Options – DLSS 2.0 offers users 3 image quality modes – Quality, Balanced, Performance – that control the game’s internal rendering resolution, with Performance mode enabling up to 4X super resolution (i.e. 1080p → 4K). This means more user choice, and even bigger performance boosts.
How can this help while gaming?
Currently DLSS 2.0 is only supported on a handful of games: Control, Deliver Us The Moon, MechWarrior 5: Mercenaries, and Wolfenstein: Youngblood.
DLSS uses the power of your RTX GPU to perform intelligence (AI) decision making about your graphics settings, allowing you to have your cake and eat it too! Max graphics fidelity aka visual quality AND higher frame rates. Surely this can’t be true?
It does seem too good to be true and DLSS is still a work in progress. It takes time to learn patterns to allow the software to work for you. Just like any other form of Ai, its generally data driven and without data to make informed decisions, you have a virtual paperweight or useless feature. This is not the case with DLSS in my experience.
When I used BF V with DLSS off, I preferred higher visual quality at the expense of frames. My settings were roughly high/ultra mix, 1440p @ 144fps (which I never saw obviously). Needless to say, I had issues maintaining over triple digit frames which is my bottom limit, 100+ or bust. Well, this lead me to reducing things like ambient occlusion and shadows. Surely at this time, RTX was indeed off.
After dealing with manually adjusting settings, I decided to enable DLSS and continue playing as normal. Do my surprise, I was seeing better frames and visuals that I was happy with. I did notice some texture resolutions were lower than I would have liked. I can assume if I left this running longer, it would be better over time but I lost interest in the game.
Conclusion
I’m excited to see what DLSS 2.0 has to offer when its released to more popular gaming titles that will be coming out. I believe DLSS and the RTX series of GPU’s have a lot to offer with this technology. I would be remiss if I didn’t say that Nvidia dropped the ball on performance promises, especially for those who are looking to DLSS to maximize everything about their experience.
More Information
https://www.nvidia.com/en-us/geforce/news/nvidia-dlss-2-0-a-big-leap-in-ai-rendering/