A few words about the engine and implementation
“Horizon Forbidden West” uses the Decima engine, a proprietary graphics and game engine developed by Guerrilla Games itself. This engine is known for its ability to create stunning visual effects and detailed, expansive game worlds, making it particularly well suited for rendering the vibrant and complex environments in Horizon Forbidden West. Decima enables a high level of detail in the game worlds, from lush landscapes to complex character models, and provides an immersive gaming experience. So it looks good without dragging the entire hardware into the colorful maelstrom.
The decision against ray tracing in the PC port of “Horizon Forbidden West” was mainly due to the enormous size of the game. The developers at Nixxes and Guerilla Games briefly considered implementing ray tracing, but decided against it due to the scalability of the game between the PlayStation 5 and PC, the extensive hours of cinematics and already impressive visual effects, and the strong artistic direction of the game, which they did not want to compromise. Another reason was the challenge of applying ray tracing to certain content such as alpha-tested trees, which is difficult even for shadows. Replacing transparent textures with modeled meshes might have led to the hopscotchy Dragon’s Dogma 2 effect faster than Aloy can say pug.
![](https://www.igorslab.de/wp-content/uploads/2024/03/005-980x505.jpg)
It was precisely this stable and resource-saving implementation that made me want to play the whole thing in my living room (which is actually just a TV and music room) with a real cinema feeling and banging sound wallpaper. And by the way: I even had fun twice during the whole thing: firstly researching what works and screwing what was there and then using and playing with it. So Easter is more or less saved, even if the weather draws me outside during the day. As something like “Horizon Forbidden Living Room”, because we only play in the evening and at night. I promise.
About DLSS or FSR, DLAA, TAA and SMAA
First a few basics: DLAA (Deep Learning Anti-Aliasing, unfortunately only works on NVIDIA’s GeForce cards), TAA (Temporal Anti-Aliasing) and SMAA (Subpixel Morphological Anti-Aliasing) are three different anti-aliasing techniques, each of which takes a different approach to improving image quality in video games and other graphical applications. However, the choice of the “best” anti-aliasing technique can also depend heavily on the specific application, personal preferences regarding image quality and the available hardware. In scenarios where image quality takes precedence over performance, DLAA or TAA may be preferred, despite their potentially higher performance impact. A summary of their differences and the impact on frame rate is certainly not a disadvantage at this point:
DLSS (Deep Learning Super Sampling) and FSR (FidelityFX Super Resolution)
The smart super sampling offered by both manufacturers is a revolutionary technology that aims to improve graphics performance by artificially increasing the game resolution. Games are first rendered at a lower resolution and then artificial intelligence and machine learning are used to upscale the images to a higher target resolution. This process makes it possible to improve the image quality without putting as much strain on the graphics card as would be the case with native rendering at a higher resolution. Both processes offer different modes that allow users to choose between better image quality and higher performance, depending on what is more important in their particular scenario.
DLAA (Deep Learning Anti-Aliasing)
DLAA uses deep neural networks to minimize staircase effects and other image defects and is an evolution of AI-based super-resolution techniques, specifically designed to smooth edges without the goal of increasing image resolution. DLAA tends to deliver very high image quality by not only smoothing edges but also enhancing details in textures, but it can be more power intensive than SMAA. However, through optimizations and the use of dedicated AI hardware (such as the Tensor Cores in NVIDIA GPUs), the performance impact can be mitigated somewhat. The exact performance impact depends heavily on the specific implementation and hardware, but compared to TAA, you should always expect around two to five percent FPS loss, which is less than it sounds.
DLSS (Deep Learning Super Sampling) and DLAA (Deep Learning Anti-Aliasing) are both technologies developed by NVIDIA and build on AI-based deep learning to improve the visual experience while gaming. However, they have different main goals and application areas, which means that they cannot be considered directly better or worse than the other, but rather in terms of how they are suited to specific needs. Nevertheless, there are some reasons why someone might prefer DLSS over DLAA, depending on their requirements and priorities. For gamers looking for both high image quality and a high refresh rate or performance on their systems, DLSS might naturally be more attractive than DLAA as it offers both.
TAA (Temporal Anti-Aliasing)
TAA uses information from previous frames to reduce staircasing effects in the current frame output and works by combining motion data with a smoothing function to smooth edges over time. TAA can smooth edges very effectively and has the added benefit of being able to incorporate motion blur, providing a smoother visual experience, but it can lead to some blurring in some situations, particularly during fast motion. TAA is generally less performance intensive than DLAA, mainly because it doesn’t require specialized hardware, but can still have a noticeable impact on performance depending on the game and implementation. Unfortunately.
SMAA (Subpixel Morphological Anti-Aliasing)
A more advanced form of Morphological Anti-Aliasing is SMAA, which recognizes and smoothes edges based on their geometric shapes, including taking subpixel details into account. SMAA offers a good balance between performance and image quality by effectively smoothing edges without adding too much blur, and among the three techniques discussed here, it tends to be the most performance-efficient option. SMAA offers a significant improvement over traditional methods such as MSAA, with a much lower performance impact.
Now let’s look at the differences in the slider and see that they are smaller than you might actually notice when playing. For DLSS and FSR, the presets for quality apply and WQHD is used for all six exceptions:
I have attached everything again as a full-screen gallery precisely because it is so difficult to see the differences, although you really have to look very closely. In FSR, the movement causes partial (but rather rare) edge flickering or slight artifacts, but at the latest when the next mecha dino appears, you no longer have any eyes free for it. However, it shows once again how well the game has been optimized and that vegetarian texture wallpapers flicker less than fully rendered solids. Long live the transparent alpha channel!
DLSS and also FSR also offer options that allow a balance between image quality and performance, giving users more control over their gaming experience. As DLSS can operate at different resolutions and performance modes, it is suitable for a wider range of hardware configurations and gaming needs. DLAA, on the other hand, is ideal for those who are already happy with their system performance and want to focus on maximizing image quality. A matter of taste, as always. And there’s also TAA and SMAA on the menu.
133 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Veteran
1
Veteran
Veteran
Urgestein
Urgestein
Mitglied
Urgestein
1
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
1
1
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →