Should you disable HPET for gaming, and does it reduce input lag?

Disabling HPET does not reduce input lag on a normal gaming PC, because Windows is not timing anything with HPET to begin with. Since Windows 7, Microsoft's own documentation says most PCs use the processor's time stamp counter (TSC) as the basis for the performance counter, and drop back to HPET only when the TSC is unsuitable. The tweak that actually costs you something runs the other way: forcing HPET on with bcdedit /set useplatformclock yes. If that flag is set, delete it. Otherwise, leave everything at default.

Why two answers circulate

HPET advice comes from two different eras, and both halves get repeated as one rule.

In Microsoft's "Acquiring high-resolution time stamps", the Windows Vista section says every computer that shipped with Vista used HPET or the ACPI PM timer as the basis for QueryPerformanceCounter (QPC), and that "such platform timers have higher access latency than the TSC". From Windows 7, "the majority" of computers use constant-rate TSCs instead, and the Windows 8 and 8.1 section says those versions "use TSCs as the basis for the performance counter". On systems where the TSC is not suitable for timekeeping, "Windows automatically selects a platform counter (either the HPET timer or the ACPI PM timer)".

So "HPET is slow" is true, and "disable HPET" sounds like the fix. But the only HPET change with a documented mechanism is removing a forced useplatformclock flag, which pushes a modern PC back onto the platform timer. Turning HPET off in the BIOS of a PC that was never using it for QPC removes nothing from your click-to-photon chain.

What the platform timer actually costs

QPC is what Microsoft tells native code to use for timestamps. The same document gives the cost of each source:

Per call, the platform timer is roughly 27 to 33 times more expensive, and Microsoft adds that a platform counter is "shared between multiple processors", which "limits scalability of QPC if it is called concurrently from multiple processors".

What the table cannot tell you is how many calls a given game makes per frame, because no game publishes that. The honest test is to measure your own latency before and after removing a forced flag, and to watch frame times and 1% lows rather than average fps.

Every HPET tweak, and what to do with it

"Disable HPET" names at least five different settings. They do different things:

The BCDEdit /set reference opens with a caution that changing some boot options "could render your computer inoperable". Every timer option on that page carries the same note: debugging only. None is described as a performance setting.

How to check your PC in two minutes

1. Confirm the TSC is invariant. Download Coreinfo from Sysinternals and run coreinfo -f. Find the line TSC-INVARIANT * TSC runs at constant rate. An asterisk means true. This is the check Microsoft's QPC document itself shows. The current release lists Windows 11 as its minimum client.
2. Look for forced flags. In an administrator command prompt, run bcdedit /enum {current}. If useplatformclock, useplatformtick, disabledynamictick or tscsyncpolicy appear, someone set them.
3. Delete them. Run bcdedit /deletevalue useplatformclock, and the same for any other listed option. Per the /deletevalue reference, you might need to suspend BitLocker and Secure Boot first, and the change applies after a restart.

Do not use the performance counter frequency as your test. Microsoft says some newer Windows versions always report 10 MHz, and warns: "don't assume that QueryPerformanceFrequency will return a value derived from the hardware frequency."

Where this stops applying

If Coreinfo shows a dash on TSC-INVARIANT, or Windows has decided your TSC cannot be synchronized across cores, Windows has already chosen a platform timer by itself, which Microsoft describes mainly for large server systems with multiple clock domains. Inside a virtual machine, Microsoft expects the hypervisor to clear the invariant TSC bit where live migration could change the TSC frequency, so a guest reading is not a reading of your hardware. And none of this applies to Arm-based Windows PCs, which have no TSC or HPET and use the Arm Generic Timer.

Frequently asked questions

Does disabling HPET reduce input lag?

Not on a default Windows install. Microsoft's documentation says Windows uses the processor's TSC as the basis for the performance counter on most systems since Windows 7, and falls back to HPET or the ACPI PM timer only when the TSC is not suitable. If HPET is not the timer in use, switching it off changes nothing in that path.

Should I turn HPET off in the BIOS?

There is no documented benefit. On a system with an invariant TSC, Windows does not use HPET for the performance counter anyway. On a system without one, Windows needs a platform timer, and HPET or the PM timer is what it falls back to. Microsoft also notes that Windows 7 and Windows 8 hardware certification requires HPET support in the platform.

What does bcdedit /set useplatformclock yes do?

Microsoft's BCDEdit reference says it "forces the use of the platform clock as the system's performance counter" and that the option "should only be used for debugging". On a TSC-capable PC it swaps a counter that is read in about 30 ns for one that costs 0.8 to 1.0 microseconds per call and needs a system call.

How do I undo useplatformclock?

From an administrator command prompt, run bcdedit /deletevalue useplatformclock and restart. Microsoft documents /deletevalue as the way to remove options that were added with /set, and a boot option change only takes effect after a restart.

Should I set disabledynamictick yes?

Microsoft describes it only as enabling and disabling the dynamic timer tick feature, and says the option "should only be used for debugging". It publishes no gaming or latency benefit for it. Leave it unset.