The total input-to-photon chain is exactly that: a chain, and 2026 end-to-end measurements consistently show that the mouse contributes under a millisecond at the USB stage, the game engine contributes the largest pre-display stage, and the display contributes the largest single stage once all variables are accounted for. Walking through each link with actual numbers, drawn from NVIDIA’s Reflex documentation, Igor’s Lab’s April 2026 latency study, and Switchblade Gaming’s Reflex-vs-Anti-Lag benchmarks, makes the practical answer obvious before any money changes hands.
What the chain actually contains
A click travels through switch debounce, USB polling, operating system input queue, game engine update, render queue, GPUGPU. Graphics Processing Unit — the chip that renders the game's visuals; the main driver of framerate and image quality. render, display scanout, and panel pixel response. Each stage adds wait time, and the totals compound. The math behind that compounding is the part most readers skip and most readers get wrong: each link is roughly independent, and the worst-case stage dominates.
The variable portion is the sum of USB wait plus the wait for the next game frame. Microsoft documents USB interrupt intervals at high-speed USB down to 125 microseconds. At 1,000 Hz polling the interval is 1.0 ms (average wait 0.5 ms); at 8,000 Hz polling the interval is 0.125 ms (average wait 0.0625 ms). The average theoretical advantage of 8K over 1K is 0.4375 ms in raw bus terms; it maxes out at 0.875 ms. The rest of the latency budget lives elsewhere.
The frame-time wait is the clock most competitive players underestimate. At 120 Hz the frame is 8.33 ms, so on average the input waits 4.17 ms for the next frame to begin; at 240 Hz the average wait drops to 2.08 ms; at 360 Hz it drops to 1.39 ms. Even at 360 Hz, frame waiting is still meaningfully larger than the 8K polling advantage. This is the single biggest reason high-refresh monitors feel different from 144 Hz panels: the average wait time per frame halves as you cross 240-360 Hz.
Where Reflex and Anti-Lag actually help
NVIDIA’s Reflex Low Latency mode targets the render queue, specifically the delay between the CPU submitting frames and the GPU rendering them. Reflex pages CPU work to land just before the GPU is ready, eliminating the queue. Reflex 2 with Frame Warp, announced at CES 2025 and requiring RTX 50-series GPUs, adds a second optimization: the CPU calculates the next camera position from the latest input and warps the rendered frame to match before sending it to the display.
The measured impact in real games is large. NVIDIA’s own THE FINALS test at 4K max settings on an RTX 5070: baseline latency was 56 ms; Reflex Low Latency cut it to 27 ms; Reflex 2 Frame Warp cut it to 14 ms. That is a 75% total reduction. In Valorant on an RTX 5090 (a CPU-bottlenecked title), Reflex 2 Frame Warp pulls average system latency under 3 ms. Reflex 2 is currently live in two titles — THE FINALS and Valorant — with support planned for older RTX cards.
AMD has two relevant technologies. Original Anti-Lag, which works at the driver level without game integration, tops out around 31% latency reduction in DirectX 9-11 titles and can actively misfire in DX12 and Vulkan games. Anti-Lag 2, which requires SDK integration like Reflex, hits 11 ms input latency in Counter-Strike 2 at Very High — a sub-15 ms result that competes with Reflex. Anti-Lag 2 is currently live in three titles: CS2, Dota 2, and Ghost of Tsushima Director’s Cut.
Igor’s Lab’s independent head-to-head between an RTX 3070 Ti and an RX 6700 XT is the clearest real-world comparison. In Valorant the cards perform comparably, with Reflex claiming only a 3 ms advantage at 4K. In Overwatch 2 the RTX 3070 Ti running at roughly 100 fpsframerate. How many images (frames) the game shows per second; higher = smoother motion. 60 fps is a common target. with Reflex held lower latency than the RX 6700 XT running at over 200 fps with Anti-Lag active. AMD’s hardware could not close the software gap the way Reflex’s engine-level integration closes it.
Display scanout is the floor you cannot move
Display scanout is where most buyers over- or under-spend. The frame does not appear all at once — the panel scans row by row from top to bottom. The scanout wait depends on refresh rate:
| Refresh Rate | Full Scanout | Average Position |
|---|---|---|
| 60 Hz | 16.67 ms | 8.33 ms |
| 120 Hz | 8.33 ms | 4.17 ms |
| 240 Hz | 4.17 ms | 2.08 ms |
| 360 Hz | 2.78 ms | 1.39 ms |
OLED pixel response is measured in tenths of a millisecond and is rarely the bottleneck at 240 Hz or above. LCD response and TV processing pipelines can add several milliseconds. On a 60 Hz TV in game mode the total adds an additional processing delay that can be larger than the scanout itself.
This is the single stage where budget most consistently buys measurable improvement. The math: a 60 Hz panel adds 8.33 ms of average scanout wait to a low-latency rig that is otherwise dialed in. A 240 Hz panel subtracts roughly 6 ms of scanout wait for the same rig. The mouse contribution does not change. The GPU work does not change. The frame-time contribution drops 2 ms. The end-to-end math: the 60 Hz panel’s 8.33 ms scanout is roughly 30% of a typical 25 ms end-to-end 60 Hz pipeline. The 240 Hz panel’s 2.08 ms is roughly 7% of a typical 30 ms 240 Hz pipeline. You do not need to spend $400 on a 4K 144 Hz OLED to feel this; a solid 240 Hz IPS at $250 already gets most of it.
How the rest of the chain measures, and what gamers get wrong
Igor’s Lab ran end-to-end measurements on two mice at 1K and 8K polling. With optical switches, raising debounce from 0 ms to 1 ms added about 0.9 ms at 1K polling and 2.0 ms at 8K polling, and the 8K-versus-1K delta measured 0.6 ms at 1 ms debounce and 1.7 ms at 0 ms debounce. The headline: switch quality and debounce tuning matter more for the mouse contribution than polling rate. In Igor’s data, a 1K mouse with optical switches and 0 ms debounce matches, and slightly beats, an 8K mouse with 1 ms debounce. The display and USB chain adds a fraction of a millisecond for wired connections and several milliseconds for wireless depending on the protocol. The sub-millisecond bus gap between wired 8K and wired 1K is mathematically real but rarely perceptible.
Without any low-latency mode, a GPU-bound game’s render queue can hold one to three frames, adding 8.33 ms at 120 Hz per queued frame, 4.17 ms at 240 Hz, 2.78 ms at 360 Hz. CPU hitches show up in 1% lows more than average FPS, which is why frame-time consistency matters more than average FPS for input lag.
The single most common overspend in 2026 is on the mouse. The 8K polling premium is mathematically capped at 0.875 ms of bus advantage, and real end-to-end measurements show 0.6-1.7 ms depending on debounce. The single most common underspend is on the monitor. A 240 Hz panel shaves an average of 4 ms of scanout wait off the chain; a 4000 Hz mouse shaves 0.4 ms of bus wait. Another common underspend is on enabling Reflex or Anti-Lag in the first place, or knowing which AMD version applies. Anti-Lag 2 is only live in three games, and original Anti-Lag should be turned off in DX12 titles where it can raise latency. The 20-70% latency reduction reported for Reflex in supported titles is real and verifiable through NVIDIA’s Reflex Analyzer on supported G-SYNC monitors.
What it means for buying decisions
The chain from finger to photon is eight stages, and the budget to actually improve it ranks roughly in this order for a 2026 buyer: panel first (240 Hz or higher, OLED if budget allows), then GPU with low-latency mode enabled (RTX 50-series with Reflex 2 if the supported games are your list; AMD RX 9070 XT with Anti-Lag 2 only in CS2, Dota 2, or GoT Director’s Cut), then CPU and memory tuned so the 1% lows hold (DDR5DDR5. A generation of system RAM (memory); newer and faster than DDR4.-6000 CL30 minimum, EXPO or XMP enabled, frame cap 3-5 fps below refresh), then mouse with optical switches and 0 ms debounce. Polling rate above 1000 Hz is the last purchase to make, and it is the one with the smallest impact.
Beyond purchase decisions, the operational fixes are free and often skipped. Dropping mouse debounce from 1 ms to 0 ms saves an average of about 0.9 ms at 1K polling and 2.0 ms at 8K polling. Setting a stable frame cap 3-5 fps below the monitor’s refresh stabilizes frame pacing and reduces VRR-induced jitter. Enabling game-mode display settings on a TV removes the TV’s processing-pipeline wait. None of these fixes cost anything, and each saves a measurable amount of end-to-end wait.