The Illusion of Detail
The prevailing advice for tuning upscalers like FSR or DLSS suggests that a sharper image is inherently a better one. This premise relies on a simple assumption: that the blurriness inherent in low-resolution rendering is the primary defect, and that sharpening is the direct cure. In practice, this view ignores the fundamental mechanism of how these algorithms reconstruct pixels. When an upscaler takes a low-resolution input and maps it to a higher-resolution output, it is not simply revealing hidden detail. It is synthesising new pixel data based on mathematical assumptions about what should exist between the sampled points. The "softness" you see is not just a lack of information; it is a deliberate smoothing of the reconstruction error. By applying a sharpening kernel, you are not unmasking truth. You are forcing the algorithm to make a stronger commitment to specific edge locations. This creates a visual state that appears crisper in isolation but is fundamentally unstable. The detail you perceive is often a hallucination of the filter, a high-frequency signal that was never present in the source data but is now amplified to the point of visibility.
This misconception persists because most visual comparisons are conducted using static screenshots. In a still frame, the amplified edges read as clarity. The human visual system is highly sensitive to local contrast, and sharpening increases this contrast, tricking the eye into registering the image as "sharper." However, this static assessment fails to account for the temporal dimension of gameplay. A screenshot is a single point in time, whereas a game is a continuous stream of changing data. When the camera moves, or when objects shift, the position of these synthesised edges changes relative to the pixel grid. If the sharpening is too aggressive, the algorithm's guess about where the edge should be becomes volatile. It swings back and forth as the sub-pixel alignment shifts. This creates a shimmering effect that is absent in static images but becomes the dominant visual artefact in motion. The "detail" you added in the settings menu is, in motion, simply noise that has been dressed up as texture.
The Physics of Ringing
To understand why this happens, one must look at the mathematics of convolution. Sharpening filters are essentially high-pass filters that subtract a blurred version of the image from the original, or they amplify high-frequency components. When this process encounters a high-contrast edge, such as a bright window against a dark sky, the filter cannot simply stop at the edge. The mathematical operation spills over, creating oscillations in the pixel values on either side of the boundary. This is known as ringing. In a photograph, this might appear as a faint halo. In a real-time rendered frame, it appears as a jagged, vibrating outline. The more you sharpen, the more pronounced these oscillations become. They are not random noise; they are a predictable byproduct of the filter trying to force a sharp transition in a space where the data is insufficient.
The critical issue is that these oscillations are highly sensitive to sub-pixel positioning. As an object moves across the screen, the exact location of the edge relative to the pixel grid changes by fractions of a pixel. A sharpening filter that is tuned for a specific alignment will produce different ringing patterns at different alignments. When the alignment shifts, the ringing pattern shifts with it. This creates a temporal instability. The edge does not just move; it vibrates. This vibration is what players perceive as shimmer. It is a direct consequence of the trade-off: you have gained spatial contrast at the cost of temporal stability. The condition under which this stops being true is when the source resolution is high enough that the edge is already well-defined across multiple pixels. In that case, the filter has enough data to work with, and the ringing is minimal. But in the context of upscaling, where the source data is sparse, the ringing is amplified exponentially with each step up in sharpening intensity.
Motion as the True Test
Most settings guides fail because they optimise for the wrong metric. They ask, "Does this look sharp?" rather than "Does this look stable?" In a static scene, a high sharpening value can look impressive. The textures appear detailed, and the lines are crisp. However, the moment the player rotates the camera or moves the character, the illusion collapses. The shimmer becomes distracting, and the image begins to look noisy rather than clear. This is why many players find that a setting that looks perfect in a menu screen becomes unusable in actual gameplay. The visual system is far more sensitive to motion artefacts than to static softness. A slightly soft image that moves smoothly is generally more comfortable to watch than a sharp image that shimmers. The brain interprets the shimmer as a defect in the rendering, whereas it interprets softness as a stylistic choice or a limitation of the resolution.
The implication for players is that the "best" sharpening setting is not the one that maximises contrast, but the one that minimises temporal instability. This requires a different approach to tuning. Instead of looking at a single frame, you must observe the image in motion. You need to look for edges that are vibrating or shimmering as they move across the screen. If you see this, the sharpening is too high. The goal is to find the threshold where the image remains stable during motion. This threshold varies by game, by scene, and by the specific upscaler algorithm. Some algorithms are more robust to motion than others, but the fundamental trade-off remains. You are always exchanging stability for contrast. The question is not whether you want more detail, but whether you can tolerate the instability that comes with it.
The Role of the Algorithm
Different upscalers handle this trade-off in different ways. FSR, being an algorithmic solution, relies on edge detection and directional filtering. It is highly responsive to the sharpening parameter, meaning that small changes can have large effects on the amount of ringing. DLSS, which uses a neural network, is trained to minimise temporal instability. It uses motion vectors and feedback from previous frames to smooth out the transitions. This makes DLSS more tolerant of higher sharpening values, but it does not eliminate the trade-off. Even with neural reconstruction, pushing the sharpening too far will introduce artefacts that the network was not trained to handle. The network is designed to produce stable, high-quality images, but it is not a magic eraser for excessive sharpening. It can mitigate the effects, but it cannot reverse them.
The key difference is in the type of artefact produced. FSR tends to produce more visible ringing on hard edges, while DLSS may produce more subtle blurring or ghosting. However, both suffer from the same fundamental problem: sharpening amplifies the uncertainty of the reconstruction. The more you sharpen, the more the algorithm is forced to guess, and the more likely it is to make a wrong guess. This is why the "best" setting is often lower than what the static image suggests. It is a compromise that prioritises the stability of the image over its perceived sharpness. Players who ignore this and crank up the sharpening are essentially asking the algorithm to lie more convincingly. The result is an image that looks great in a screenshot but falls apart in the real-time environment.
Tuning for the Eye, Not the Screen
The final step in understanding this trade-off is to recognise that the goal is not to maximise technical metrics, but to optimise for human perception. The human eye is not a perfect sensor. It has its own biases and limitations. It is sensitive to high contrast, but it is also sensitive to motion. A setting that looks "sharp" on a monitor may actually be causing visual fatigue due to the shimmer. The best setting is the one that is invisible. It is the setting where you do not notice the upscaling, where the edges are stable, and where the image feels solid. This often means accepting a certain amount of softness. It means recognising that the "detail" you are chasing is an illusion, and that the stability of the image is more important than its clarity.
To achieve this, players should adopt a dynamic tuning process. Start with the default setting. Then, increase the sharpening in small increments. After each change, move the camera or look at a moving object. Watch for shimmer. If you see it, back off. Repeat until you find the highest setting that does not produce visible motion artefacts. This will likely be lower than what you thought was "sharp." But it will be the setting that allows you to play the game for hours without your eyes aching. It is the setting that respects the physical limitations of the upscaling process. It is the setting that acknowledges that sharpening is not a free lunch. It is a trade-off, and like all trade-offs, it requires a careful balance. The goal is not to have the sharpest image, but the most stable one.