125 shares, 338 points

NVIDIA recently introduced Image Scaling as a competitor to AMD’s FidelityFX Super Resolution technology. This technology of the green team will work on display cards of AMD and Intel as well as NVIDIA display cards.

On the other hand, an open source SDK (software development kit) has been released so that this technology can be used in games and applications. In other words, anyone who wants will be able to use Image Scaling in their games and applications.

NVIDIA also says that DLSS technology has differences from other scaling technologies. In this article, we will take a closer look at DLSS and the differences between other scaling technologies.

NVIDIA often uses the term spatial amplifiers (dimensional scalers) when talking about the details and is actually referring to AMD’s FSR technology in this regard. In addition, the newly introduced Image Scaling feature works similarly to AMD’s FSR technology.

NVIDIA DLSS, Image Scaling and AMD FSR

There are profound differences between NVIDIA DLSS and spatial amplifiers (scalers) such as NVIDIA Image Scaling, AMD FSR, and bilinear (bi-linear), bicubic (bicubic), and Lanczos filters. These key differences between DLSS and spatial amplifiers will become clearer when we explain the advantages of DLSS and the limitations of other technologies.

DLSS is a technique designed from the ground up to overcome scaling limitations using supersampling, temporal feedback and artificial intelligence techniques to provide image quality comparable to native resolutions with significant performance gains. DLSS manages to offer better detail, more stable images and edge quality than spatial amplifiers.

Spatial amplifiers lack these technologies to create native quality images and trade IQ for performance. Without supersampling-trained AI techniques, they cannot deliver image quality equivalent to native resolutions, and images look much worse at higher scaling factors and lower resolutions. They also lack additional motion-based artifacts, as they lack temporal feedback.

Generally speaking, the difference is that spatial amplifiers use older and simpler techniques, sacrificing image quality for performance. In this context, DLSS uses powerful new techniques designed to preserve image quality.

The images below of the Necromunda game show the image quality differences and all the details that the two technologies can produce. DLSS can use the lowest quality setting, Performance, to achieve the spatial amplifier’s image quality at the highest quality setting, Ultra Quality. NVIDIA’s advanced technology can offer additional high performance as well as high visuality, as it preserves image quality.

FSR Ultra Quality on the left; DLSS Performance on the right; AMD Ryzen 9 5950X processor, 32GB 3333MHz DDR4 memory, RTX 3060 GPU.

When comparing the image quality between DLSS and spatial amplifiers, it is possible to say that the technology supported by Tensor cores provides superior quality in terms of fine details even at lower settings.

Another Necromunda example below highlights that the texts are much clearer and more detailed in the game using DLSS. It’s also worth noting that this destruction of fine detail is common to all spatial amplifiers, as prescaled resolutions initially lack this detail. DLSS uses artificial intelligence and supersampling techniques to preserve details.

The left spatial amplifier fails to render text or other fine details correctly. DLSS, on the other hand, is able to achieve this despite using a more aggressive upscale here and having far fewer pixels to work with.

DLSS Advantages over Spatial Amplifiers

  • True reconstructed high resolution detail
  • Continuously learning artificial intelligence model
  • Accuracy close to real images
  • Thinner lines
  • Clearer texts
  • Reducing movementollary objects
  • Compensating flicker
  • Prevent sharpening of objects
  • More quality with fewer pixels

Advantages of DLSS

DLSS also has higher quality inputs than spatial amplifiers and uses information from multiple frames, just like shooting a long exposure movie. Different examples of how objects should look in each frame are collected, and this information is motion vectors to help track objects in motion, and artificial intelligence is used to best understand the data. Artificial intelligence also combines these frameworks so that the processing is more complete and successful in terms of what information it will retrieve from previous frameworks.

In this example, when considering 1440p Quality mode, after collecting more than 6 million pixels of information, 3.5 million pixels are output. In addition, NVIDIA has an artificial intelligence network, and with the inputs, high image quality can be presented comparable to the native image. Let’s also note that in some cases higher-than-local details are produced.

By contrast, spatial amplifiers use a fixed algorithm for scaling, yielding a single frame with a lower resolution than native (in this example, 2.2 million pixels in Ultra Quality mode) and less information. Therefore, this relatively simple technique used offers lower performance compared to native image quality, and it becomes inevitable to compromise on image quality for performance gains.

DLSS provides more detail as the artificial intelligence network (called convolutional autoencoder) is trained on large datasets of 16,000-resolution images. DLSS then uses its trained AI network to learn how to generate a high-resolution supersampled output sample from a lower resolution frame. So much so that this process works pixel by pixel. This technology knows what the upgraded visuals will look like, what they should look like, and it does a good job at it.

Spatial amplifiers, on the other hand, sample low-resolution pixels at just one spot, then upscale and sharpen the image. Sharpening cannot create additional detail from low resolution information and can only increase the local contrast of details already available at lower resolution. Not only does it use less data than local, but it goes through a fixed function algorithm to upgrade using less information than local. As a result, we sacrifice image quality for performance in this process.

Another advantage found in DLSS is motion awareness. The green team’s technique uses motion vectors of motiondollarsy objects as well as temporal feedback from previous frames to reduce typical motion objects generated by scaling. Spatial amplifiers, on the other hand, lack temporal feedback, which makes them incomplete when it comes to motion instability, jitters, and paddling objects.

DLSS is a future-proof scaling technology with an always-learning artificial model. It is also constantly being improved through continuing education on NVIDIA supercomputers. With each major release of DLSS, better image quality has been achieved along with a wider range of games/applications. Some developers prefer to update old games to take advantage of these improvements. At the time of writing this article, DLSS 2.3 version has just been released. With the new version, DLSS SDK 2.3 was also released, which enables even smarter use of motion vectors to reduce ghosting and improve image quality.

Technology that leverages deep learning continues to improve games like Cyberpunk 2077, reducing ghosting and improving image quality. Made on Tesdollarser Ultra/High Quality Preset, 1080p DLSS Quality mode, RTX 3060, Ryzen 5950X, 32GB RAM.

Another example is the Control game. Improved visual artifacts, fine detail, and object movement in fan wings when updated with the game DLSS.

NVIDIA Driver Based Spatial Amplifier – Image Scaling

Players have actually been using the old-school spatial upgrade for a long time, and many games have scaling details in their settings options. Originally released in 2019, NVIDIA Image Scaling can be enabled via the NVIDIA Control Panel and GeForce Experience.

NVIDIA says Image Scaling has been updated with a new algorithm that uses 4-way scaling and adaptive sharpening filters and a 6-touch filter to improve performance. Sharpening and scaling are done in one pass, thus highlighting that it is very efficient compared to existing spatial algorithms. It can also be used in all games and has a sharpening option accessible through the GeForce Experience layer for instant customization.

NVIDIA Image Scaling and DLSS Difference

The updated and newly introduced feature is mostly only available at target resolutions of 4K and 1440p with low scaling factors, like a spatial amplifier and other spatial amplifiers. It also doesn’t do as well as DLSS when it comes to motion objects and has limitations like other spatial amplifiers.

It works at image resolutions from 1080p to 4K compared to DLSS, NVIDIA Image Scaling and other spatial amplifiers. Therefore, DLSS has its place and it makes more sense to compare Image Scaling technology with other scaling technologies such as AMD FidelityFX Super Resolution.

The Godfall 4K comparison below shows the similarity between NVIDIA Image Scaling and FSR. FSR Ultra Quality 4K on the left; On the right, 77% at 4K scaling, NVIDIA Image Scaling active at default sharpening. Additionally, tesdollarser was built with AMD Ryzen 9 3900X processor, 32GB 3200MHz DDR4 memory and RTX 3080 Ti graphics card.

This 4K Resident Evil 8 comparison shows that NVIDIA Image Scaling looks sharper than a similar spatial amplifier. While spatial amplifiers all suffer from the same kind of issues compared to a superior temporal scaling AI technology like DLSS, they can look a little different from each other depending on the exact filters they use and the scenario.

Resident Evil 8: FSR Ultra Quality 4K left; 77% at right 4K scaling, NVIDIA Image Scaling at default sharpening; AMD Ryzen 9 3900X, 32GB 3200MHz DDR4, RTX 3080 Ti.

Image Quality Issues

The following list of image quality is related to image scaling technologies. The images on the left show the problem, examples of which are described.


Unstable edges and lines seem to move or glow. Moire patterns are also included in this category. You can look at the finer details (right) versus the blur at the ends of the hair (left).


Duplicate images or edges as objects move due to user input or game animation. We can describe this phenomenon, known as ghosting, as an echo, as in a second image or sound that follows the object. You can detect ghosting by focusing on an object moving in the background that has a different value or hue. You can better understand the difference if you choose a high contrast object, such as a dark tree against a bright sky, or a spinning fan blade. If you’re playing a multiplayer game, you can also focus on players moving remotely across the map.

Fan canadollaryellow (left) looks like a ghost behind them as they move, while harekedollarsi fan canadollaryellow (right) leave no image trail.


Note exactly overlapping edges at the edges, especially on objects that are not nearly perfectly horizontal or vertical to the screen. Better to have the game in motion to see the difference.

The edges on the left are jagged, while the edges on the right retain anti-aliasing.


Less detail in the interior of the objects, or typically diffused into high-detail objects. You need to compare closely with local images. Test both stationary and in motion e. Because lower resolution textures are used, blurring can often be the result of game-side mipmap bias errors.

Blur: The image on the left is blurry, the crosswalk on the right is sharp.

amik Detail in Content

It can be defined as the minimization of details in rapidly changing content. You can carefully look at the series of rain, snow, blown sand, dust particles, digital signage text and other particle effects. You can also focus on smoke, fog, fire, padlocks, gun barrels, and anything else that moves fast in games like a range of other gun effects.

The particle effect details (embers and flame fragments) that should be in the left image are missing.

Detail Removed

Observe telephone hadollar yellows, antennas on top of buildings, radio towers, chidollar series and other similar objects with fine geometry, these details are greatly reduced or completely absent with other technologies. DLSS helps preserve these fine details, especially when viewing these objects from afar.

Fine railing and plant details (left) are either missing or undersized. It is also more distorted compared to the natural (right) image.

UI Breakdown

See HUD, game navigation markers, zoomed gun sight, holographic sights, and the like. These issues can be easy to find, as the UI is often static, always present, and full of straight lines. You may see issues such as flickering, ghosting, and chromatic aberration/color channel separation in the UI.

Ghosting can be seen in the on-screen HUD reticle (left).

Detail on Reflections

Detail aberration on reflected surfaces. Look for reflected shapes and objects on reflective surfaces and observe for noisy or jagged edges.

Reflective detail is rough and noisy (left) compared to normal (right).

Shadow Edges

Rough, pixelated, or low quality. Do some objects have lower quality shadows than others? Look for aliasing or pitted edges. Note the border area where shadows are reduced and how technology affects this gradation e.

Shadows have jagged edges (left) compared to local (right).


An object that has been squashed, pixelated, or abstracted. Shadows are more common in amic and detailed content such as fire and smoke.

Flames appear pixelated (left).


Anty, grainy pattern, as if film grain has been added. These distortions are especially visible when moving. It can also be a byproduct of oversharpening.

Ant particles (left) appear almost as an atmospheric haze, especially when in motion.


Strange behavior with very bright values ​​like reflections in broken glass. Look for specular highlights and bright lights and evaluate if they are rendered correctly.

Note the overemphasis (left) compared to the local image (right) e.

Sharpened Artifacts

Edge distortion issues related to oversharpening. Oversharpening can cause artifacts at edges such as halos where local contrast can produce light areas around the edges, distorted and jagged edges, and generally noisy grain. Excessive sharpening of enlarged images also results in the “merging” of colors, creating a painting-like type of image. For more details and examples, see the rest of the article.

Note: You may not be able to stumble upon all problems throughout the game. Some may be of short duration or you may only encounter them a few times. A common problem throughout the game can be more annoying.

Quality Benchmark

Here are some points to consider when comparing performance improvements to image quality.

DLSS and spatial amplifiers are fundamentally different technologies, so their image quality will not match exactly. Balance the settings to the point where both technologies look as similar as possible. As a general rule, DLSS generally provides similar image quality with the Performance setting as spatial amplifiers do at 4K with Ultra Quality scaling rates of 77%.

Evaluating Image Nedollarsic

Spatial upscaling techniques rely on the sharpening process to clean up blurry, raised textures to give a sense of higher resolution. While nedollarification is subjective to users, many people tend to have a bias for sharpened images.

Sharpening is really just increased edge contrast. When an image is sharpened, fine edge gradients are reduced, and this is called acutans. When viewed from the correct viewing distance, our eyes, which are very sensitive to local contrast, interpret sharpness as enhanced detail or resolution. That’s why a little sharpening may seem appealing to gamers.

Artificial Objects and Tingling (Parazidollarsenme)

The first issue is with the limitation: sharpening requires sufficient resolution to be of benefit. In other words, it cannot develop details and can only act on the object that is there. That’s why 1080p/1440p resolutions and large scaling factors aren’t very successful.

The second problem is that sharpening can damage it and make the image look worse. If the image lacks detail, as with upgraded images, sharpening can cause distortions in the newly upgraded resolution.

As for the Myst example below, the spatial amplification sharpening further distorts the detail of the paddle wheels and also creates a parasidollary grain area, especially in the varying hues of the water. Oversharpening can often cause problems in areas with smooth gradients, such as sky, water, and solid color areas. However, DLSS does not rely solely on the sharpening technique and offers true high-resolution details free of these issues.

Myst 1080p: spatial scaler creates artifacts when trying to highlight edges that were not initially defined enough due to the low resolution. Increased sharpening contrast creates tingling in smooth water gradients. FSR Ultra Quality on the left; DLSS Quality on the right; AMD Ryzen 9 5950X, 32GB 3200MHz DDR4, RTX 3080.


Another type of damage that excessive sharpening can cause is called ringing. Ringing is a type of oversharpening that causes a lighter peripheral halo around the sharpened edges. This is easily visible on dark edges against a lighter background. In the unscaled native 4K Myst example below, the aggressively used sharpening algorithm creates ringing issues where rocks meet the sky and other background areas.

Loss of Detail

Because sharpening works by reducing brightness gradients near the edges, the resolution may initially appear higher because our perceptual systems are sensitive to contrast. However, while the image resolution does not improve, it does not stay at the same level and decreases.

In another Myst example, the oversharpened version increases the darkest tones and increases the amount of bright tones while the midtones are compressed on both sides. Not only does this result in an overall chalky (like chalky) appearance with an exaggerated amount of brightness, but it can also make the image look dull. The native version is richer with more color information and gradients, as seen in the average swatches.

Myst 4K: Extreme in-game sharpening has been applied (left) in this unscaled 4K benchmark. The end result was an overall image with less saturation and color, with exaggerated darks and lights. Average color samples represent the “chalky” quality that some sharpened images have. Exaggerated sharpening algorithm at 100% on the left; local game on the right; both tested with Core i9 10900K CPU, 32GB 3200MHz DDR4 memory and RTX 3080 Ti graphics card.

Excessive nedollarization of scaled images can combine areas of solid color with “pictorial” deposits (such as muddy water). The effect is cumulative throughout the scene and can create an undesirable cartoon art style as these flat, blocky areas resemble traditional comic book and cartoon animation processes. The example below shows how DLSS maintains its intended realistic art style even at scaled 1080p, while spatial scalers can change the art type and aesthetic intent.

Myst 1080p: The spatial scaler on the left changes the game’s art style to a more cartoonish version of the original art style. On the right (DLSS) the intended realistic art style is preserved. FSR Ultra Quality on the left; DLSS Quality on the right; both tested on the same system: AMD Ryzen 9 5950X, 32GB 3200MHz DDR4, RTX 3080.

These are just a few of the problems with oversharpened games. A bit of sharpening may be appreciated by gamers. However, this applied sharpening can be lost in some scenarios due to image distortion, additional artifacts, loss of color fidelity. However, these solutions can also give good results when applied correctly according to the game and visuals.

Intel Front and XeSS (Xe Super Sampling)

The Blue team is planning to make great strides in every field with the Arc series screen cards yellow designed for high-performance gaming. As it is known, software is as important as hardware in this field. We know that the new display cards with the code name “Alchemist” have all the capabilities of DirectX 12 Ultimate, including ray tracing.

Intel announced its own XeSS (Xe SuperSampling) technology in response to NVIDIA DLSS and AMD FidelityFX Super Resolution solutions. Intel’s boost technology will use the XMX instruction found on DG2 GPUs (Arc Alchemist) or the DP4a instruction set found in competing products such as NVIDIA Pascal GA102-106 GPUs, depending on the moment.

Intel XeSS technology falls into the same category as NVIDIA DLSS technology, as it is based on artificial intelligence and neural networks. This technology will work with motion vectors and history buffering, which should improve the quality of graphics in motion.

Looking specifically at the presentation made with The Riftbreaker game, the XeSS upscaled 4K image clearly looks much sharper than a standard upscaled image, although the frame rates don’t look like much.

Intel also ran XeSS in Hitman 3 and provided some visuals.

You can find the demo of Rift Breaker running on Intel Arc GPU right below. With the transition from native 1080P resolution to XeSS 4K, the visuals literally change. While comparing the two side by side, the game, in which XeSS technology is used, draws attention with much sharper edges and higher quality textures.

XeSS does this by using machine learning to reconstruct subpixels. Intel’s technology is similar to the approach to upscaling with DLSS, and as you know, NVIDIA’s tensor cores use artificial intelligence or machine learning to enhance the image.

The most important difference of XeSS is that it does not require hardware-accelerated AI cores to work. Intel will also use open standard yellow for this solution to work on NVIDIA and AMD GPUs.

Like it? Share with your friends!

125 shares, 338 points

What's Your Reaction?

hate hate
confused confused
fail fail
fun fun
geeky geeky
love love
lol lol
omg omg
win win
Michael Lewis



Your email address will not be published. Required fields are marked *

Choose A Format
Personality quiz
Series of questions that intends to reveal something about the personality
Trivia quiz
Series of questions with right and wrong answers that intends to check knowledge
Voting to make decisions or determine opinions
Formatted Text with Embeds and Visuals
The Classic Internet Listicles
The Classic Internet Countdowns
Open List
Submit your own item and vote up for the best submission
Ranked List
Upvote or downvote to decide the best list item
Upload your own images to make custom memes
Youtube, Vimeo or Vine Embeds
Soundcloud or Mixcloud Embeds
Photo or GIF
GIF format