Frame generation vs FPS: what the extra frames actually change

Frame generation can make motion look smoother and push an FPS counter higher. It does this by creating intermediate images from frames the game rendered and motion information. Those extra images are not additional full frames rendered by the game, so a display showing 120 images per second does not by itself mean the game is rendering 120 complete frames each second. NVIDIA and AMD describe this process in their DLSS and FSR documentation.

That difference matters when you choose a graphics setting or compare performance numbers. Frame generation can improve the visual rhythm of motion, but the number on screen does not tell you whether controls feel more immediate.

What an extra frame actually is

In a simple mode that inserts one image between each pair of game-rendered frames, 60 rendered frames could become about 120 presented images per second. AMD used this as an ideal FSR 3 example, not a guaranteed result. Actual output depends on the game, implementation, base frame rate, hardware and settings.

The generated image is estimated from neighbouring game frames and movement data. It is not a new, independently rendered game state with a fresh input update. That is why generated frames can make a turn or camera pan look smoother without making the game behave exactly like it was rendering every frame itself.

The estimate also has to account for moving objects, effects and interface elements. NVIDIA’s September 2022 DLSS 3 explanation describes combining optical flow with engine motion vectors; its example notes that relying on motion vectors alone can leave a shadow appearing to stutter. That illustrates why generated imagery can show artifacts, but it does not predict what every game will show.

Frame generation is different from upscaling

These settings can appear together in the same menu, but they do different jobs:

• Upscaling renders a game frame at a lower internal resolution and reconstructs it at the output resolution. It may raise the rate of game-rendered frames because fewer pixels need to be drawn.
• Frame generation estimates one or more in-between images from game-rendered frames. It can raise the number of images presented to the display without raising the number of full frames the game rendered by the same amount.
• Latency reduction changes parts of the CPU, GPU or frame-presentation pipeline to reduce the delay between an input and its on-screen response. NVIDIA Reflex and AMD latency-reduction features are separate technologies that may accompany frame generation in supported games and hardware.

If responsiveness is your priority, try upscaling on its own first when the game offers it. That can improve the base frame rate without adding interpolated images. Whether the result looks good still depends on the game and quality setting.

Why the FPS number and the feel can disagree

Three things are easy to mix up: the rate of game-rendered frames, the final rate of images presented, and how quickly the game responds to your input.

A comparison is useful only if you know what was counted. If a game provides separate base and presented rates, record both. Otherwise, use the same FPS counter for both runs and treat its number as a performance indicator, not a latency measurement.

Latency can also change. AMD’s September 2023 FSR 3 developer guidance says interpolation adds latency by design and describes techniques intended to reduce its impact. That is vendor guidance about FSR 3, not a measured result for every game. The net feel depends on the implementation, base frame rate, upscaling and the rest of the system.

AMD’s guidance is specific about its FSR 3 implementation: it recommends at least 60 FPS after upscaling and before interpolation. Treat that as a vendor recommendation for FSR 3, not a universal minimum for every game or frame-generation system.

When frame generation is worth trying

Frame generation is most useful when the game already runs at a reasonably steady base rate and you want smoother motion. It may be less appealing in a competitive game where small changes in control response matter, or when visual artifacts are distracting.

It is a graphics preference, not a fix for a recurring hitch or a guarantee that a low base rate will feel responsive. If motion looks worse around fast-moving objects, the HUD or scene changes, compare the same moments with the feature off.

Test the trade-off in your game

Use a repeatable built-in benchmark, saved game or short route. Keep the game version, resolution, graphics preset, upscaler and other settings the same.

1. Start with frame generation off. Play the route and note the base frame rate, whether motion is steady, and how aiming or camera control feels.
2. Turn on frame generation only. Repeat the same route. Use the same FPS counter; if the game exposes both base and presented rates, record both.
3. Check the picture and the controls separately. Look for ghosting, flicker or unstable edges during motion, then judge whether input still feels responsive. A larger FPS number is only one part of the result.
4. Test a latency option separately if the game supports one. Change one setting at a time so you can tell what affected the result.
5. Keep the setting only if it helps the experience you want. If the image is smoother but control feels worse, turn frame generation off or compare upscaling on its own.

This is a documentation-based guide, not a Jitsumi benchmark. The useful result is the one you observe in the game and hardware you actually use.

Sources and further reading

• NVIDIA: Introducing DLSS 3 — frame generation inputs, intermediate images and Reflex integration.
• AMD GPUOpen: FSR 3 game integrations and technical details — the 60 FPS recommendation is specific to AMD FSR 3, with latency and implementation caveats.