How to reduce input lag on Windows 11.

01What input lag actually is

When you click, the event travels through a chain: the mouse samples it, USB delivers it, Windows routes it, the game simulates a new frame, the CPU hands draw calls to the GPU, the GPU renders, the finished frame waits in a queue, and the display finally draws it. Every link adds time, and the total — usually somewhere between roughly 20 and 100 ms on a typical setup — is what your hands feel.

You can't change the speed of light or your monitor's panel, but you can shorten the middle of the chain: how long frames wait in queues, how often the system checks for input, and how quickly the CPU and GPU respond after idle. That's what everything below targets.

One framing rule before we start: higher framerate is the biggest lever. Each frame rendered at 60 fps occupies about 16.7 ms of the chain by itself. If a setting change buys you more frames per second, it usually buys you lower latency too — the two goals mostly point the same way.

02Fullscreen, borderless, and the flip model

Older Windows versions had a real gap between "exclusive fullscreen" (fast) and "borderless windowed" (an extra compositor copy). Windows 11 narrows it: modern games on the DXGI flip model can bypass the compositor even in borderless mode, so the old reflex of forcing exclusive fullscreen everywhere is no longer automatically right.

• Per-game, not global. Test both modes in the games you actually play. Some older titles still benefit clearly from exclusive fullscreen.
• "Optimizations for windowed games" (Settings → System → Display → Graphics) moves eligible older DX10/11 games onto the flip model. Generally worth turning on.
• Disable per-game fullscreen optimizations only when a specific game misbehaves — it's a compatibility escape hatch, not a speed switch.

03Hardware-accelerated GPU scheduling

HAGS (Settings → System → Display → Graphics → Default graphics settings) hands frame scheduling from the Windows kernel to the GPU itself. On recent NVIDIA and AMD cards with current drivers it can shave a little latency and CPU overhead — and it is required for some latency features like NVIDIA's frame-generation pipeline.

The honest caveat: results vary by GPU generation and driver version. A small number of setups see stutter with it on. Treat it as a default-on candidate that you verify, not a guaranteed win.

04Keep the render queue short

The largest avoidable chunk of input lag on most systems is the render queue: finished CPU frames waiting for a busy GPU. When the GPU runs at 99-100% load, the queue fills, and your clicks describe a world several frames old.

• Use the vendor latency mode. NVIDIA Reflex (in supported games) or Ultra Low Latency in the driver; Radeon Anti-Lag on AMD. These exist precisely to keep the queue at or near one frame.
• Cap your framerate slightly below the ceiling. An in-game or driver cap a few fps under what your GPU can sustain keeps GPU load out of the saturation zone — frames stop queueing. With a VRR display (G-SYNC / FreeSync), cap a few fps below the panel's refresh rate for the same reason.
• V-Sync without VRR is a latency tax. If tearing bothers you, prefer VRR + a frame cap over plain V-Sync.

05Mouse settings that matter

• Turn off "Enhance pointer precision." It's acceleration — it changes how far the cursor moves depending on speed. Most games read raw input anyway, but the desktop setting still leaks into some titles and into your muscle memory.
• Polling rate: 1000 Hz is a sane default on a gaming mouse (a 1 ms sampling interval vs 8 ms at 125 Hz). Higher rates exist; the gains shrink fast and can cost CPU in some games.
• Use raw input in-game when the option exists — it bypasses Windows pointer processing entirely.

06Power plans and timer resolution

Two system-level switches change how fast the machine reacts after idle moments:

• Power plan. Balanced plans park cores and drop clocks aggressively; the first milliseconds after an input can be spent waking hardware up. A high-performance plan (or a tuned minimum processor state) trades watts for responsiveness. On laptops, mind the heat budget — sustained boost can throttle and cost more than it buys.
• Timer resolution. Windows schedules work on a system timer that apps can raise from the default 15.6 ms granularity to 0.5-1 ms. Most games request this themselves; since Windows 10 2004 the request is per-process and Windows 11 manages it more strictly. Forcing it globally helps mainly with older titles and certain frame limiters — a "depends on your game" tweak, not a universal one.

07What to skip

• Registry "mouse fix" packs from forum posts — most date to Windows 7 curve tables and do nothing on a raw-input game except risk your settings.
• Disabling every Windows service you don't recognize. A handful of background processes matter (see the measuring section); bulk service-killing mostly breaks things.
• "Ping boosters" for local lag. Network latency and input latency are different chains. No DNS change or traffic tool shortens the path between your mouse and your GPU.
• Anything that promises a fixed number. Latency gains depend on your display, GPU load, and the game's engine. Distrust any tool quoting a universal "-XX ms".

08Measure, change one thing, measure again

You can't feel 3 ms, but you can measure trends. A frametime overlay (RivaTuner Statistics Server — RTSS) shows whether your GPU is saturated and whether a change made frame delivery steadier. NVIDIA's overlay can show full render latency on Reflex games. Change one setting at a time, play five real minutes, compare. Keep what measures better, revert what doesn't.