The Illusion of Native Resolution
The prevailing advice for tuning FidelityFX Super Resolution or NVIDIA DLSS is deceptively simple: if the image looks soft, increase the sharpness. This intuition stems from a fundamental misunderstanding of what upscaling actually achieves. When a game renders at a lower internal resolution, such as 1080p on a 4K display, the system must synthesize the missing pixels. This process inherently blurs high-frequency details because there is simply no data to define them. The sharpness slider does not recover lost information; it amplifies existing contrast to trick the human visual system into perceiving edges where none strictly exist. The result is a subjective improvement in clarity, but it is a manufactured one. Players often interpret this heightened edge contrast as a return to native resolution quality, when in reality they are merely observing a more aggressive manipulation of the pixel data. The danger lies in the assumption that "sharper" equals "better." In visual fidelity terms, it is merely a different kind of error. The goal of upscaling is not to make the image look sharp, but to make it look correct. By prioritising the former, users inadvertently introduce a second layer of visual noise that compounds the original resolution deficit. This creates a false positive in the quality assessment, where the image appears more detailed than it actually is, masking the underlying limitations of the lower render target.
The Physics of Ringing
To understand why excessive sharpening is detrimental, one must examine the specific artefact it produces: ringing. When a sharpening kernel operates on a high-contrast edge, such as a bright window frame against a dark wall, it does not just enhance the boundary. It overshoots the target contrast, creating a halo of inverted colour or brightness adjacent to the edge. This phenomenon is mathematically similar to the Gibbs phenomenon in signal processing, where approximating a discontinuity with a finite number of terms results in oscillations near the jump. In the context of a video game, this manifests as faint, ghostly lines or halos around distinct objects. In a static image, these halos can be subtle and easily overlooked, often blending into the background or being perceived as a stylistic choice. However, the mechanism is destructive. The kernel is effectively adding energy to the signal that was not present in the source data. This added energy is not random noise; it is structured, predictable, and directly correlated with the position of high-contrast edges. Because it is structured, the human eye is particularly sensitive to it. The brain interprets these halos as a violation of physical plausibility, even if the conscious mind registers the image as "crisp." The trade-off is precise: you gain immediate, localised contrast at the cost of introducing global, structured artefacts. The more you sharpen, the more pronounced these halos become, until they begin to interfere with the perception of the scene's geometry. This is the cost of the illusion, and it is paid in every frame where a high-contrast edge is present.
Motion Exposes the Flaw
The critical omission in most settings guides is the difference between a still image and a moving scene. When the camera pans or the player moves, the position of high-contrast edges changes relative to the pixel grid. In a static frame, the ringing halos are stationary. They sit in the same pixel locations from one frame to the next, allowing the eye to adapt to them. They become part of the static texture of the image. However, in motion, these halos move with the edges. As the edge shifts, the overshoot shifts with it, creating a temporal inconsistency. The eye perceives this as shimmer or jitter. A bright edge that is slightly too sharp will appear to vibrate or flicker as it traverses the screen. This is because the ringing artefact is not a stable property of the object; it is a dynamic property of the upscaling process. The amplitude of the halo changes as the edge aligns differently with the sub-pixel sampling grid. This temporal instability is far more distracting than the initial softness of the unsharpened image. A soft image is stable; it is consistent across frames. A heavily sharpened image is unstable; it is a source of constant, low-level visual interference. The condition under which this stops being true is when the motion is extremely slow or when the edges are low-contrast. In those cases, the temporal variation is minimal, and the shimmer is imperceptible. But in fast-paced gameplay, where high-contrast edges are moving rapidly, the shimmer becomes a significant source of eye strain and visual fatigue.
Tuning for the Eye, Not the Screenshot
Most players tune their upscaler settings by taking a screenshot or pausing the game. This is a fundamentally flawed methodology. A screenshot freezes the temporal dimension, removing the very factor that makes excessive sharpening problematic. In a paused state, the ringing halos are static and can be ignored by the brain. The image looks clean, detailed, and professional. This leads to the conclusion that the setting is optimal. However, once the game resumes, the temporal instability returns. The setting that looked perfect in the pause menu becomes a source of visual noise in actual play. The correct approach is to tune the sharpness slider while the game is in motion. Observe a scene with high-contrast edges moving across the screen, such as a bright building passing by a dark tree. Watch for shimmer. If you see a faint, oscillating glow around the edges, you have sharpened too much. Reduce the setting until the shimmer disappears. The resulting image may look slightly softer than the maximised setting, but it will be temporally stable. This stability is what allows the eye to focus on the gameplay rather than the artefacts. The goal is not to maximise the perceived detail in a single frame, but to minimise the visual interference across a sequence of frames. This is a subtle but crucial distinction. It shifts the focus from a spatial metric (sharpness) to a temporal metric (stability). By tuning for motion, you are acknowledging that the human visual system processes video, not stills. You are accepting that a slightly softer image is better than a sharp, shimmering one. This is the true cost of upscaling: the compromise between spatial clarity and temporal consistency.
The Verdict on the Slider
The claim that more sharpening makes upscaled images look better is only true under a very specific, and often irrelevant, condition: when the image is static and the viewer is not sensitive to temporal artefacts. For the vast majority of gaming scenarios, this condition does not hold. The mechanism of sharpening is a double-edged sword. It provides a quick, easy win in perceived clarity, but it does so by introducing a structured artefact that becomes a liability in motion. The ringing halos are not a minor cosmetic issue; they are a fundamental flaw in the signal processing that degrades the viewing experience over time. The reason existing coverage misses this is that it is difficult to demonstrate in a written guide or a static image. You cannot show shimmer in a photograph. You have to feel it in your eyes. This is why the advice has remained stuck in the past, treating upscaling as a static image problem rather than a dynamic video one. The player’s role is to act as their own quality control system. Do not trust the screenshot. Trust the motion. Turn the sharpness down until the world stops vibrating. That is the point where you have stopped fighting the upscaler and started working with it. The result will be an image that is less aggressively detailed but far more comfortable to look at. It is a trade-off that most players are not aware they are making, but it is the one that determines whether the technology enhances the game or distracts from it. The slider is not a volume knob for detail; it is a dial for stability. Treat it accordingly, and the upscaled image will finally look like what it is supposed to be: a high-quality approximation of the native render, not a noisy, shimmering imitation.