Upscaling and Frame Generation: How to Use DLSS, FSR, XeSS and Frame Gen

This guide is for players who see DLSS, FSR, XeSS and "frame generation" in a settings menu and are not sure how to use them, and for anyone who has noticed a requirements table saying a tier assumes upscaling and wondered what that implies. We cover how upscaling and frame generation work, what the quality modes mean, why frame generation affects input latency, and where latency-reduction technologies such as Reflex and Anti-Lag fit in.

In the early days, the only way to make a demanding game playable was to lower the resolution and switch off effects, and the image suffered visibly. Modern upscaling offers another route: render at a lower resolution and reconstruct an image close to the higher one. Frame generation goes a step further by inserting computed frames between rendered ones. Both are genuinely useful, and both involve trade-offs, so it pays to understand how they work.

What upscaling does

An upscaler renders the game at a lower internal resolution and scales it up to your display resolution. Fewer pixels to compute means less GPU work. The difference between approaches is how they scale:

• Spatial upscaling uses only the current frame. It is cheap and widely compatible, but recovers limited detail.
• Temporal upscaling combines previous frames and motion vectors to accumulate detail. It usually looks clearly better than spatial methods and anti-aliases at the same time.
• Machine-learning upscaling builds on temporal methods with a trained model that decides how to reconstruct detail, and usually needs specific hardware support.

The three main technologies

Support changes with new versions and drivers, so treat the game menu and the GPU vendor's official documentation as the final word on what is available.

Choosing a quality mode

Upscalers offer several modes, which differ in the ratio between internal and output resolution. Typical names and approximate ratios are below; exact values can vary by technology and version:

Some rules of thumb:

• The higher the output resolution, the lower the mode you can tolerate. At 4K, Performance mode still renders a fairly high internal resolution; at 1080p the internal image is already small, and blur or artifacts become more likely.
• At 1080p, start with Quality and try Balanced only if needed.
• The same mode can look different in different games, because integration quality and the accuracy of motion vectors affect the result.

Common upscaling artifacts and related settings

Upscaling is not free, and knowing the typical artifacts tells you which way to adjust:

• Ghosting and trails. Temporal methods rely on previous frames, so fast-moving objects, or inaccurate motion vectors supplied by the game, can leave faint trails behind objects.
• Shimmering fine detail. Power lines, fences and hair can flicker or break up when the internal resolution is low.
• Softness. Some games look slightly soft after upscaling and offer a sharpening slider. Too much sharpening creates halos and noise, so adjust it sparingly.

Some games also offer dynamic resolution, which adjusts internal resolution on the fly to hold a target frame rate. It reduces drops in heavy scenes, but image clarity changes from moment to moment; if you prefer consistent quality, use a fixed quality mode instead.

When artifacts bother you, try another quality mode or a different upscaler, because each technology can behave quite differently from one game to the next. It is worth spending a few minutes in a demanding scene comparing the options side by side rather than settling on the first one that appears in the menu.

How frame generation works

Frame generation computes one or more new frames between two rendered ones, raising the frame rate shown on screen. The result looks smoother, but there are important conditions.

Why it adds latency

To insert a frame between two rendered frames, the system has to wait for the second one to finish before it can compute the one in between, so displaying real frames is delayed slightly. On top of that, generated frames do not reflect new input; your actions appear only in the next rendered frame. Frame generation improves how smooth the game looks, not how quickly it responds.

Base frame rate matters

The quality and feel of frame generation depend heavily on the frame rate before it is enabled:

• At a higher base frame rate, consecutive frames differ less, generated frames are more accurate and the added latency is less noticeable.
• At a low base frame rate, frames differ more, so ghosting and broken UI text become likelier and the latency is easier to feel.

That is why vendors generally recommend getting reasonably smooth performance without frame generation first, then using it to add smoothness, rather than using it to rescue a game that does not run well.

Where it fits and where it does not

• Good fits: single-player and slower-paced games, or making motion smoother on a high-refresh monitor.
• Poor fits: reaction-critical competitive games, very low base frame rates, and cards already short on VRAM, since frame generation needs extra memory.

Reflex, Anti-Lag and latency reduction

System latency is the time from pressing a key to seeing the result. When the GPU is fully loaded, frames prepared by the CPU queue up waiting to be rendered, which adds latency. Latency-reduction technologies synchronize CPU and GPU more tightly to shorten that queue:

• NVIDIA Reflex requires game integration and is commonly paired with DLSS frame generation to offset some of its added latency.
• AMD Anti-Lag exists as a driver-level feature and as a newer version that requires game integration.
• Intel Xe Low Latency is Intel's counterpart, paired with XeSS frame generation.

Even without frame generation, enabling these in supported competitive games usually lowers latency.

Driver-level frame generation

Besides in-game frame generation, some vendors offer driver-level versions, such as AMD Fluid Motion Frames. These work without game integration, but because they lack the game's motion vectors and UI information, quality and consistency are usually below in-game implementations, and HUD elements are more prone to artifacts.

A practical sequence

1. Check baseline performance and the bottleneck at native resolution.
2. If the GPU is the limit, enable upscaling, starting with Quality.
3. If that is not enough, try Balanced or Performance, or lower expensive settings as described in our graphics settings guide.
4. Once the base experience is acceptable and you want more smoothness, consider frame generation, along with the matching latency-reduction feature.
5. If the CPU is the bottleneck, upscaling helps little, and frame generation raises displayed frame rate without fixing CPU-side stutter. See our CPU guide.

Common misconceptions

• "Generated frames count the same as real ones." They do not respond to input; responsiveness still follows the base frame rate.
• "Upscaling always looks blurry." Temporal and ML upscalers in Quality mode often look close to native and can improve anti-aliasing.
• "Frame generation lets a weak card run anything." At low base frame rates, both artifacts and latency become obvious.
• "Every upscaler behaves the same on every card." Hardware paths, versions and game integration all affect results.
• "The frame rate in the requirements is native performance." If a tier assumes upscaling or frame generation, native performance may differ; see how to read system requirements.

How this relates to FrameScope

More and more games note in their official requirements that a tier assumes upscaling. Can I Run It places you using the published requirements and our GPU hierarchy, but does not promise frame rates with any particular technology enabled, because the result depends on the game, the technology version and the driver. See our methodology.