Input lag budget calculator
Total input lag is not one number, it is a chain of six. This calculator estimates each link from your actual settings, ranks them, and names the one worth fixing. Almost nobody's worst link is the one they were about to spend money on.
Your worst link is -.
| Stage | Cost | Share | Relative |
|---|
The chain, in order
When you press a mouse button, six things have to happen before a photon leaves your monitor carrying the result. Each has a cost, and the costs are wildly different in size. Understanding the ordering is what stops you spending four hundred pounds on the fourth-largest term.
| Stage | Typical cost | What it actually is | The lever |
|---|---|---|---|
| Human reaction (not part of the machine) | 180 to 250 ms | Simple visual reaction time. Everything below is what you add on top. | - |
| Switch debounce (mechanical mouse switch) | 0 to 8 ms | Firmware waits to confirm the contact is stable. Optical switches skip this. | Optical or low-debounce firmware |
| USB / wireless report wait | 0.06 to 4 ms | Half a polling interval on average. 1 kHz = 0.5 ms. | Raise polling rate |
| Wireless radio hop | 0.2 to 1 ms | Modern 2.4 GHz dongles. Bluetooth is 10x worse and unusable competitively. | Use the dongle, never Bluetooth |
| OS and game input sampling | 0 to 1 frame | The game reads input once per frame at a fixed point in its loop. | Higher framerate |
| CPU simulation + render queue | 2 to 30 ms | The biggest single variable. A deep queue at a GPU bottleneck dominates everything else. | Reflex / Anti-Lag, framerate cap below GPU limit |
| GPU render | 2 to 16 ms | Roughly one frame time when GPU bound. | Lower settings, higher fps |
| Scan-out wait | 0 to 1 frame time | The frame waits for the display to start drawing. Mean is half a frame time. | Higher refresh rate |
| Panel response (grey to grey) | 0.03 to 12 ms | OLED is effectively instant; slow VA can smear past the next frame. | OLED, or a well-tuned overdrive |
| Pixel to perceived brightness | 1 to 8 ms | Sample-and-hold blur, not a fixed delay, but it moves where you perceive the edge. | Backlight strobing, higher refresh |
What the arithmetic tells you
The queue is usually the villain
Every other term in the chain is bounded by hardware physics and lands somewhere between zero and a few milliseconds. The render queue is bounded by nothing except your settings, and at a GPU bottleneck it can reach thirty. If you change one thing after reading this page, cap your framerate a little below what your GPU can sustain, or enable Nvidia Reflex, AMD Anti-Lag or the equivalent in-engine option. It costs nothing and it is frequently worth more than every peripheral upgrade you have considered.
Polling rate matters enormously, then suddenly does not
Moving from 125 Hz to 1000 Hz removes 3.5 ms of mean delay. That is a bigger saving than upgrading from a 144 Hz display to a 240 Hz one (which saves 1.4 ms of scan-out wait). Moving from 1000 Hz to 8000 Hz removes 0.44 ms, which is roughly a tenth of the first jump for many times the cost.
| Polling rate | Report interval | Mean added delay | Saved vs previous | Verdict |
|---|---|---|---|---|
| 125 Hz | 8.000 ms | 4.000 ms | - | Audibly dated. Fix this first. |
| 250 Hz | 4.000 ms | 2.000 ms | -2.000 ms | Audibly dated. Fix this first. |
| 500 Hz | 2.000 ms | 1.000 ms | -1.000 ms | Fine in practice, one step from optimal. |
| 1000 Hz | 1.000 ms | 0.500 ms | -0.500 ms | The point of diminishing returns for almost everyone. |
| 2000 Hz | 0.500 ms | 0.250 ms | -0.250 ms | Measurable, barely perceivable. |
| 4000 Hz | 0.250 ms | 0.125 ms | -0.125 ms | Sub-0.13 ms saving. CPU cost starts to matter. |
| 8000 Hz | 0.125 ms | 0.063 ms | -0.063 ms | Marginal gain, real CPU overhead on weaker systems. |
Refresh rate buys two things, and one of them is not latency
A higher refresh rate reduces the mean scan-out wait, which is genuinely part of your latency. But the larger benefit is motion clarity: more unique images per second means a moving target smears less across your retina, so you can see it while it moves rather than only when it stops. That is a perceptual benefit, not a latency one, and it is why 240 Hz feels transformative in a way the 1.4 ms saving does not explain. See motion clarity.
The fix order, from this calculator's output
- Cap framerate or enable Reflex to collapse the render queue. Free.
- Set polling to 1000 Hz. Free.
- Raise framerate, which shrinks two terms at once (input sampling and queue). Costs settings quality or hardware.
- Raise refresh rate, once framerate can feed it. Expensive but real.
- Everything else. Switch type and panel response are typically under 5 ms combined on modern gear.
Related reading
- The click-to-photon chain in full, with what each stage physically is.
- Reflex and Anti-Lag: what they do to the queue and when they do nothing.
- How to measure your own latency with a phone camera.
- Network latency, which is a separate budget and often a larger one.
Frequently asked questions
Are these numbers exact for my machine?
No, and no calculator can be. The polling, frame timing and scan-out terms are exact arithmetic from the values you enter. The debounce, queue depth and panel response terms are estimates you select, because measuring them requires a high-speed camera or a hardware latency tester. The value of the output is the ranking, not the total: it tells you which link dominates, which is the decision you are actually making.
Why is human reaction time not included?
Because it is not something you can buy your way out of, and including it makes every hardware difference look trivial by comparison. Simple visual reaction time is roughly 180 to 250 ms. That does not mean 10 ms of system latency is irrelevant: reaction time is a fixed offset applied to everyone, while system latency is a variable you control, and in a duel where both players react at the same moment the one whose bullet is registered first wins.
What is the render queue and why does it dominate?
To keep the GPU fed, the CPU prepares frames ahead of time and stacks them in a queue. When the GPU is the bottleneck, that queue fills up, and every frame in it is a frame that was simulated before your input existed. A full three-frame queue at 60 fps is 50 ms of pure staleness, which is larger than every peripheral term combined. This is why Reflex and Anti-Lag exist and why they are the single highest-value toggle in the list.
Is 8000 Hz polling worth it?
Arithmetically it removes 0.44 ms of mean delay compared with 1000 Hz. That is real but small, and it costs measurable CPU time on weaker systems, which can cost you more in frame pacing than it saves in report timing. Get to 1000 Hz, which removes 3.5 ms from a 125 Hz baseline, and treat everything above it as a rounding error you may take if it is free.
How do I actually measure my own latency?
Three ways, in increasing order of accuracy: an in-game overlay such as Nvidia's PC latency readout, which covers CPU to render but not the peripheral or display; a high-speed phone camera at 240 fps filming your mouse and screen together, which is crude but end to end; or a dedicated hardware latency tester. Our measuring latency guide walks through the camera method, which costs nothing.