When the NVIDIA GeForce RTX cards were initially launched almost two years ago with real-time hardware raytracing, people were skeptical about the whole thing being a gimmick. While achieving stable framerates with real-time raytracing turned on was quite hard at first, NVIDIA’s machine learning technology called DLSS quickly made it a viable feature.
Deep Learning Super Sampling or DLSS uses machine learning to boost framerates in games at higher resolutions while also trying to not sacrifice the image quality too much. This has been extremely handy when it comes to providing playable framerates with RTX enabled on the lower-end RTX graphics cards. As it is based on machine learning, the feature has continued to improve over time with better optimization for specific games.
However, with the new NVIDIA DLSS 2.0, the graphics card manufacturer has promised some massive gains both in performance and image quality. In a blog post published today, NVIDIA called DLSS 2.0 a “big leap in AI rendering” and highlighted the key features of the new update that separate it from 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.
The biggest improvement with DLSS 2.0 is arguably the amount of flexibility that it offers in terms of hardware. Where DLSS 1.0 fell apart at times was its ability to provide better performance at lower resolutions. However, the new update makes things much more viable across the board. In addition to that, the new improved AI network uses Tensor cores much more efficiently than its predecessor making it twice as fast.
Another big improvement in DLSS 2.0 is the ability to upscale an image by 4 times as compared to 2x on DLSS 1.0. What this means is that DLSS 2.0 is now capable of producing a 4K image from a mere 1080p source fed into it. This is done by a number of things combined together including temporal feedback, which gathers data over time and analyzes it.
One of the biggest complains that people had with DLSS 1.0 was the fact that it was not available for every single game out there since it required NVIDIA to manually train their AI for that specific game. This resulted in only a handful of big games that could take advantage of the new technology. However, DLSS 2.0 brings a ‘fully synthetic training set’ that essentially eradicates the requirement of training each game separately. This means that we can expect many more titles to integrate DLSS in the future.
The change has instantly borne fruit as in addition to games like Wolfenstein Youngblood and Deliver Us The Moon (which have already been updated to DLSS 2.0), other games have started to get the new update as well. For instance, one of the game of the year contenders last year, Control, has been updated by Remedy to DLSS 2.0.
NVIDIA also discussed better DLSS integration with games made in Unreal Engine 4 and as a result, another game that is getting DLSS 2.0 immediately is MechWarrior 5 (made in Unreal Engine 4).
The blog post also included performance comparisons with the newer version of DLSS managing to achieve a performance increase of over 30% in multiple games. You can check all of that right here.
AMD’s RDNA 2-based graphics cards with hardware raytracing capabilities are also coming out later this year and as we’ve seen from Xbox Series X and PlayStation 5 reveal, they’re no slouches either. When you pair that with Microsoft’s recent DirectX 12 ultimate integration, it seems like we are finally heading towards making raytracing a mainstream feature in most games.
With DLSS 2.0 and its excellent performance improvements, NVIDIA has paved the way for that to become a reality much sooner than we had anticipated. In addition to that, team green is also gearing up to launch its RTX 3000 series cards that will quite possibly turn hardware raytracing into a child’s play at high framerates.
Are you convinced about NVIDIA’s direction heading forward or are you still waiting to see what AMD will bring to the table? Let us know down in the comments below!