Games & Performance
Frame Generation: When It Helps and When It Hurts Latency
Frame generation can multiply FPS, but at a latency cost, learn how it works, when it helps, and when to leave it off for fast competitive play.
Games & Performance
Frame generation can multiply FPS, but at a latency cost, learn how it works, when it helps, and when to leave it off for fast competitive play.
Frame generation is one of the most misunderstood features on modern GPUs. On paper it doubles or triples your FPS counter, which sounds like free performance, but the number in the corner of your screen and the way your game actually feels are two very different things. After running dozens of benchmark passes across single-player showcases and twitchy shooters, I've landed on a simple rule: frame generation is a smoothness tool, not a responsiveness tool, and knowing that difference tells you exactly when to switch it on.
Traditional rendering pipelines produce every frame you see by running the full game logic and rasterization for each one. Frame generation changes that math. Instead of rendering every frame, the GPU renders a pair of "real" frames and then synthesizes an intermediate frame between them using motion vectors, optical flow data, and an AI model that predicts how pixels move.
The practical result is that for every real frame your GPU renders, you get one (or with multi-frame variants, several) additional inserted frames. Your monitor displays more frames per second, so motion looks visibly smoother, especially on high-refresh panels.
Here's the part the marketing slides gloss over: the generated frame contains no new information from the game engine. It is an interpolation, not a fresh simulation. Your inputs, enemy positions, and physics are not sampled again for those inserted frames. They are visual filler, beautifully convincing filler, but filler nonetheless.
This is the counterintuitive core of the whole topic. You'd think more frames means a more responsive game. It's the opposite, and the reason is structural.
To insert a frame between two real frames, the GPU has to already have the "next" real frame in hand before it can show you the in-between one. That means the pipeline deliberately holds a finished frame back for a moment so it has both endpoints to interpolate from. That held-back frame is latency you can feel in your hands.
So the trade looks like this:
A game running at a real 60 FPS that gets pushed to a displayed 120 FPS with frame generation still only responds to you 60 times a second. It just looks like 120.
Both major vendors pair frame generation with a low-latency mode (NVIDIA Reflex, AMD's Anti-Lag equivalents) precisely because they know generation adds lag. These features trim the render queue and reduce the CPU-to-GPU backlog, clawing back some of the latency the generation step introduces.
The important nuance: turning on Reflex-style latency reduction with frame generation usually gets you back to roughly where you started, latency-wise, not below it. It's a mitigation, not a cure. If a game offers the low-latency toggle, always enable it alongside frame generation. Leaving it off is the worst of both worlds.
If you take one thing away, make it this. Frame generation quality is almost entirely determined by your base frame rate before generation kicks in.
The reason is straightforward once you think about the interpolation. The lower your real frame rate, the further apart your two endpoint frames are in time, and the more the AI has to guess about what happened in between. Wider gaps mean more visible artifacts and, crucially, worse latency because each held-back frame represents a longer stretch of real time.
My working guidelines from testing:
The trap people fall into is using frame generation to rescue an unplayable frame rate. That's exactly the scenario where it performs worst. It is a multiplier for an already-decent experience, not a floor-raiser for a bad one.
I keep it enabled in a specific class of games, and I'm happy to when the conditions fit.
In these cases the smoothness is real and pleasant, and honestly, once you're locked into a controller for a slower-paced game, the added latency sits well under the threshold where most people notice it.
The flip side is just as clear-cut.
Beyond latency, watch for visual giveaways that generation is struggling. The clearest ones I look for:
If you spot these, your base frame rate is too low or the game is a poor fit. That's your signal to disable the feature rather than push through it.
You don't need lab equipment to make a good call. My quick evaluation routine for any game:
If the base rate was healthy and the game is slower-paced, you'll usually keep it on. If your hands hesitate or the base rate was already marginal, trust that instinct over the frame counter.
Frame generation is a legitimately great feature that has been oversold as free performance. It is not free, and it is not performance in the sense that matters for aiming and reacting. It's a motion-smoothness enhancer that costs you a little latency, and it delivers its best value in exactly one scenario: a single-player or slower-paced game that's already running at a solid real frame rate on a high-refresh display.
Use it there and enjoy it. Keep it off in competitive shooters and anywhere your base frame rate is struggling. And whatever the FPS counter says, let the way the game feels in your hands make the final call, that number was never telling the whole story anyway.
Keep reading
Squeeze more frames from your current PC for free, driver tweaks, settings, and system fixes that boost in-game FPS without buying new hardware.
DLSS, FSR, and XeSS compared head to head, learn how each upscaler works, where image quality differs, and which delivers the best FPS on your GPU.