Background Raw Input listeners in Windows 10 and Windows 11

On this page

Short answer: the background listener cap of roughly 125 Hz is confirmed by our physical measurement on Windows 11 24H2. With system throttling enabled, the average interval of background WM_INPUT was 7.97 ms, or about 125.5 Hz; foreground retained 1.04 ms. After disabling the mechanism, background returned to 1.00 ms. In separate virtual runs, throttling did not lead to raw packet loss: 32 of 32 and 256 of 256 events were preserved.

Status: the throttling and coalescing mechanism for background listeners is documented by Microsoft. The frequency of about 125 Hz was measured by a separate public tester with a physical 1000 Hz mouse on Windows 11 24H2. The system branch was also observed in Windows 11 25H2 and was not found in the matched Windows 10 22H2.

Why Microsoft added the cap

Microsoft states the reason directly: a high report rate mouse sent input not only to the game but also to several background processes. Processing these requests consumed noticeable CPU time that could have been spent on rendering, and substantial stutters were observed on a test Surface Laptop Studio with a 1000 Hz mouse. The solution was throttling, coalescing and a message rate cap specifically for background Raw Input listeners.

By the time the change shipped, fast mice had already gone well beyond 1000 Hz. For example, Razer released an 8000 Hz wired mouse in 2021 and 4000 Hz wireless technology in 2022. An 8000 Hz device can send up to eight times more reports per second than a 1000 Hz device. Therefore the spread of 4000–8000 Hz mice logically increased the scale of the problem with multiple background listeners.

The last sentence is our interpretation of the context, not a statement by Microsoft. Microsoft did not call the update an emergency reaction specifically to 4000 or 8000 Hz mice, and in the published test it used a 1000 Hz mouse. It is also more accurate to speak of the overall cost of delivering and processing input requests rather than attributing the entire effect to hardware interrupts alone.

This material relates to the “Reduce the rate of background Raw Input events” setting. It limits background listeners specifically and should not be described as a limit on foreground input or as a guaranteed FPS gain.

Testable claim

We tested five claims:

Windows 11 has separate handling for high-frequency background Raw Input listeners.
It is absent in the same form in the Windows 10 22H2 we studied.
On Windows 11 24H2, the effective rate of a background listener really is about 125 Hz with an input stream of about 1000 Hz.
Throttling can reduce the integrity of the game’s WM_INPUT stream through losses, merging or splitting of packets.
Changing the window mode by itself creates a different Raw Input path.

Scope of research

Windows 10 22H2 build 19045.6456;
Windows 11 24H2 with a physical 1000 Hz mouse;
Windows 11 25H2;
isolated virtual machines;
foreground and background consumers;
regular Raw Input registration, RIDEV_NOLEGACY and RIDEV_INPUTSINK;
windowed, borderless and confirmed exclusive presentation;
control runs of 32 events each and a separate run of 256 events.

The physical run checked intervals between WM_INPUT, but did not include a real match, anti-cheat or overlay. The virtual run did not reproduce USB polling, a physical GPU, a display or the click-to-photon path.

Methodology

In a separate public experiment, RawMouseThrottleBufferTester registered a mouse with RIDEV_INPUTSINK and measured Stopwatch intervals between WM_INPUT motion messages. The same window was compared in foreground and background under default system values, then after disabling throttling.

The average value shown is calculated over a ring window of the last 512 received intervals. Zero movements and pauses of 40 ms or more were discarded. The Samples field in the screenshot shows the total number of received intervals at the time of the capture, not the size of the statistical window.

Additionally, system components of Windows 10 22H2 and Windows 11 25H2 were compared statically to find a separate background mouse handling branch and to separate the Raw Input path from legacy cursor and presentation paths.

Then, in identical virtual environments, a controlled sequence of mouse events was fed to a foreground or background consumer. For each scenario, the number of sent and received raw packets, losses, merges, splits, foreground state and separately legacy/cursor branch events were recorded.

Only one factor was changed at a time: the consumer registration method, foreground state, window mode or system throttling profile. Between scenarios, the test state was returned to the recorded baseline.

Results

Physical mouse, Windows 11 24H2

State	Average interval of the last 512 events	Equivalent frequency	Samples at capture
Default, foreground	1.04 ms	≈962 Hz	2 221
Default, background	7.97 ms	≈125.5 Hz	3 556
Throttling off, foreground	1.00 ms	≈1000 Hz	19 606
Throttling off, background	1.00 ms	≈1000 Hz	12 009

This confirms about 125 Hz specifically for the background RIDEV_INPUTSINK consumer in the Windows 11 24H2 we studied. The foreground path of the same program was not capped to 125 Hz.

Controlled virtual scenarios

Scenario	Windows 10 22H2	Windows 11 25H2	Result
Baseline Raw Input delivery	32 sent, 32 received	32 sent, 32 received	No losses, merge or split detected
Background without `RIDEV_INPUTSINK`	0 of 32	0 of 32	Background delivery not requested
Background with `RIDEV_INPUTSINK`	32 of 32	32 of 32	Background delivery works on both OSes
Windowed, borderless, exclusive	32 of 32 in each mode	32 of 32 in each mode	Presentation mode did not change packet integrity
Stress throttling profiles	Not applicable	256 of 256 in all states	The legacy/cursor branch changed, but not the integrity of `WM_INPUT`

RIDEV_INPUTSINK is a documented toggle for background delivery. Without it, a background consumer should not receive the same stream as a foreground application. The zero result in this row is not Windows data loss.

In the stress run, system throttling states noticeably changed the number of legacy events and system cursor movement. At the same time, in all states the Raw Input consumer received the same 256 packets out of 256. Therefore the effect found cannot correctly be described as “Windows 11 loses Raw Input”.

What is confirmed

Microsoft added throttling, coalescing and a message rate cap for background raw mouse listeners in Windows 11.
On the Windows 11 24H2 we studied, a physical foreground consumer received messages at an interval of about 1 ms, while a background consumer received them at an interval of 7.97 ms, or about 125.5 Hz.
After disabling throttling, the background consumer’s interval returned to 1.00 ms.
In the Windows 11 25H2 we studied, a separate branch of this handling is observed; in the exact Windows 10 22H2 pair it was not detected.
RIDEV_INPUTSINK changes background delivery of WM_INPUT on both studied OSes.
In all the listed scenarios, raw packet integrity was preserved 1:1.
Throttling changes manifested in the measured legacy/cursor branch, not as raw packet loss.

What is not confirmed

That every background listener on every Windows 11 build is always capped to exactly 125.0 Hz. The confirmed result applies to the described Windows 11 24H2 and registration method.
That the mechanism always reduces FPS, latency or stutter on any computer.
That disabling system throttling improves mouse control.
That DWM manages the integrity of WM_INPUT in all games and Windows 11 builds.
That identical packet integrity guarantees identical physical click-to-photon latency or a subjective aiming feel.
That a virtual machine result carries over to every physical mouse, game, anti-cheat or overlay.

Limitations

The public physical captures do not contain the exact Windows 11 24H2 build number, the mouse model, a CSV of all intervals or an automated order of state switching. The average value reflects the last 512 events, and mouse movement was performed manually. Therefore the result confidently confirms the observed cluster around 8 ms, but does not set an exact constant for any system.

A virtual machine allows the software route to be reproduced, but does not reproduce USB polling, the mouse microcontroller, a physical GPU, a display or the full game loop. Runs of 32 and 256 events are sufficient to verify the observed integrity of a specific route, but not to estimate rare losses with low probability.

Windows 11 25H2 was compared with one exact Windows 10 22H2 build. The result should not automatically be carried over to early Windows 11, Windows Server or future updates.

For how to reproduce the dynamic part of the observations, see How to check it yourself.

BoosterX develops GameModeX and ProcessX, and this research and its tools, including the public RawMouseThrottleBufferTester, belong to the BoosterX developer, so it has a direct interest in the results. The methodology and scope of applicability are described above, and the conclusions can be verified against open data: the tool’s public code, measurement captures and the listed sources. The zero result for WM_INPUT losses, the confirmation of about 125 Hz and the absence of a universal guarantee are published together.

Practical takeaway

On Windows 11, leave the system throttling of background raw mouse listeners at its default state. Microsoft introduced it to reduce the work of background applications when using a high report rate mouse, while preserving precise foreground game input.

For a gaming PC, BoosterX recommends limiting compatible background listeners to about 50 Hz. BoosterX has not yet published its own measurement of this interval; the number itself is consistent with public independent measurements: according to checks by PC-Tuning and Noverse, an interval of about 20 ms corresponds to a compatible listener rate of about 50–60 Hz. These materials are cited in the sources as additional cross-referencing, while the value measured in this article is the system cap of about 125 Hz, not the rate after manual tuning. Increasing the interval reduces the number of delivered background events and handler invocations during mouse movement. A foreground window on the verified path retains full-speed input.

The direction of local optimization is confirmed: reducing the delivery rate of background events reduces both the number of such deliveries and the number of handler invocations. The resulting change in overall CPU load, FPS or frametime for an arbitrary set of programs was not measured. An application’s background response to the mouse may become less smooth, so a listener that genuinely needs a high rate in the background is a reason to restore the Windows default. The practical description and the exact registry state are given on the “Reduce the rate of background Raw Input events” page.

If a specific background application causes stutters or an input conflict, first update or close that application. Do not disable the system optimization or suspend processes without a reproducible comparison.

The legacy background listener limiting feature in GameModeX was intended primarily for Windows 10 and is not a replacement for the Windows 11 system mechanism. For a new configuration, the Windows 11 and ProcessX supported approach is recommended.

Restoring state

The research was performed in isolated virtual environments. Changed test states were returned to the recorded baseline between scenarios; after completion, the original virtual machine state was used. On a user’s computer, this article does not recommend changing system parameters, so no separate restore action is required.

Public primary sources

RawMouseThrottleBufferTester — a public program, source code, static observation, WinDbg check and four captures of the physical Windows 11 24H2 measurement.
Default, foreground and default, background — 1.04 ms versus 7.97 ms.
Throttling off, foreground and throttling off, background — 1.00 ms each.
Microsoft: Reduced game stutter with high report rate mice — throttling, coalescing and a cap for background raw mouse listeners.
Microsoft: KB5027303, OS build 22621.1928 — the first stable preview update with an improvement for high report rate mice.
Microsoft: Windows 11 Insider Preview Build 23424 — an early public description of the high report rate mouse improvement during gaming.
Razer: Viper 8KHz release — the official announcement of the 8000 Hz mouse dated January 28, 2021 and a comparison of report volume with 1000 Hz.
Razer: the evolution of 4000 and 8000 Hz — the official timeline of wired and wireless high polling rate devices.
PC-Tuning: background listener interval check — a public methodology for observing the interval and the RawMouseThrottleDuration range.
Noverse: independent check of two background listeners — at an interval of 20 ms, a compatible listener was measured at about 60 Hz, while a listener with bypass registration retained about 1000 Hz.
Microsoft: RAWINPUTDEVICE — the purpose of RIDEV_INPUTSINK and RIDEV_NOLEGACY.
Microsoft: About Raw Input — the registration and delivery model of WM_INPUT.

Public sources and wording verified: 2026-08-24.

Change history

2026-09-20: the recommendation of about 50 Hz was reworded: the number is explicitly cross-referenced with public independent measurements, the absence of our own published interval measurement is stated; the conflict-of-interest disclaimer was supplemented with the ownership of the research and tools, and a link to self-verification was added to the methodology.
2026-08-25: 50 Hz recommended for the gaming scenario as a confirmed reduction of background processing; the boundary for the numerical effect on overall CPU and FPS was preserved separately.
2026-08-24: added the documented CPU load context, the boundary of the conclusion about 4000–8000 Hz mice and the cautious scenario of manually limiting background listeners to about 50 Hz.
2026-08-24: added the public physical Windows 11 24H2 measurement confirming about 125 Hz for a background listener; the boundary was preserved that this is not a universal constant for every build and registration.
2026-08-24: published the first comparison of Windows 10 22H2 and Windows 11 25H2.