DLSS Frame Generation and Input Lag: The Trade-Off Explained

Frame generation multiplies the displayed frame rate by inserting synthesized frames that the GPU did not actually render from the game engine. The catch is that the synthesized frames introduce a delay between your input and what appears on screen. Understanding the mechanism determines when this trade-off is worth accepting.

DLSS Frame Generation (DLSS-G), available on RTX 40 series and newer NVIDIA GPUs, analyzes motion vectors and optical flow data between two real rendered frames and generates an interpolated frame to place between them. If the game renders at 60 FPS and frame generation is enabled, the display receives approximately 120 frames per second, with every other frame synthesized rather than rendered by the game engine.

AMD's equivalent technology, AMD Fluid Motion Frames (AFMF) and FSR 3 Frame Generation, uses a similar optical flow approach but without NVIDIA's dedicated optical flow accelerator hardware. Intel's XeSS Frame Generation serves the Arc ecosystem on the same principle.

Why Frame Generation Increases Latency

A conventional render pipeline submits a frame to the display immediately after the GPU finishes drawing it. With frame generation active, the display pipeline must hold the first real frame in a buffer while the synthesis engine generates the interpolated frame to insert after it. This holding period adds one frame of delay before the real frame reaches the display, on top of the display's own scan-out latency.

The exact latency increase depends on the frame rate before generation is enabled. At 60 FPS source frames, each frame represents approximately 16.7 ms. Holding one frame while synthesis occurs adds roughly that amount to end-to-end latency. At 90 FPS source, each frame is approximately 11 ms, reducing the added latency proportionally. This is why NVIDIA documents frame generation as intended for scenarios where the GPU is already exceeding 60 FPS without it.

NVIDIA Reflex and Latency Mitigation

NVIDIA ships DLSS Frame Generation alongside Reflex, which reduces latency in the CPU-to-GPU submission pipeline by tightening the render-ahead queue. Reflex reduces latency before the frame generation step. When both are enabled simultaneously, the net latency increase from frame generation is partially offset by Reflex's reduction elsewhere in the pipeline.

In practice, a game running at 90 FPS with Reflex enabled may measure 30–35 ms end-to-end latency. The same game with frame generation added typically measures 35–45 ms. The displayed frame rate increases from 90 to approximately 175 FPS. Whether the smoothness improvement outweighs the latency increase depends entirely on the type of game and player sensitivity.

Latency Trade-Off at Different Source Frame Rates

When Frame Generation Makes Sense

The technology is well-suited to graphically intensive games that are GPU-limited and have inherently higher latency tolerance—single-player titles, open-world games, and cinematic experiences where you are not reacting to moment-to-moment threats. In these scenarios, the visual smoothness improvement from doubling the displayed frame rate is perceptible and valuable, while the additional 6–15 ms of latency has no practical consequence.

When to Disable It

Competitive multiplayer games demand minimal latency. The input-to-pixel delay is a meaningful variable in first-person shooters, fighting games, and any title where reaction time is competitive. In these scenarios, the visual benefit of frame generation is outweighed by the latency penalty. Disable frame generation in OBS-style competitive settings and rely on rendering higher native frame rates instead, even if that means reducing graphics settings to achieve it.

Frame generation also produces visible artifacts around fast-moving objects. Hands, muzzle flashes, and rapidly moving environmental elements in the synthesized frame can exhibit ghosting or reconstruction errors. This is an image quality concern independent of the latency argument.