The High Precision Event Timer (HPET) is one of the oldest entries in the PC tweaking playbook — half the internet swears disabling it adds frames, the other half says it does nothing or hurts. The truth is nuanced and hardware-dependent. This guide explains what HPET is, when toggling it helps, and how to test it properly instead of copying someone else’s setting.
What HPET is
HPET is a hardware timer that Windows can use to schedule events precisely. Modern CPUs also expose other timers (the invariant TSC and the ACPI timer). Windows normally picks the best available timer automatically. The controversy is that on some systems forcing HPET on, or leaving a manual “useplatformclock” flag set, adds timer overhead that shows up as slightly higher DPC latency and marginally lower 1% lows.
The real-world impact
| Scenario | Typical effect |
|---|---|
| Default (Windows auto-selects) | Best for most modern systems |
| Forced HPET on via BCDEdit | Can add latency on some rigs |
| HPET disabled in BIOS | Usually neutral; occasionally smoother |
The key mistake to avoid
Most “disable HPET” guides actually tell you to run bcdedit /set useplatformclock true, which forces HPET on — the opposite of what people think. If someone set that flag chasing a tweak, the fix is to remove it: bcdedit /deletevalue useplatformclock and reboot so Windows returns to automatic timer selection. That single cleanup resolves most timer-related stutter reports.
How to test correctly
Change one thing, then measure. Note your average FPS and 1% lows in a repeatable benchmark run, and check DPC latency in LatencyMon. Toggle HPET in BIOS or clear the useplatformclock flag, reboot, and run the identical test. If numbers do not move outside run-to-run variance (usually 2-3%), the setting is irrelevant on your hardware — leave it at default.
Should most people bother?
On current AMD and Intel platforms, the safe answer is: leave Windows on automatic timer selection and make sure no manual useplatformclock flag is lingering. Only pursue HPET tweaks if LatencyMon shows timer-related overhead or you have a specific, measurable stutter you have already tried to fix with drivers and frame pacing.
The broader timer picture
HPET is only one of several timers, and Windows has become much smarter about choosing between them. On modern platforms the invariant TSC is fast and reliable, and the operating system leans on it automatically. That is why blanket “disable HPET for free FPS” advice from years ago rarely applies to current hardware — the conditions that made it matter have largely been engineered away.
When it genuinely matters
The tweak still has teeth in one situation: a system where someone previously forced a platform-clock flag, or where a specific driver interacts badly with a timer. If LatencyMon shows timer-related overhead, or you inherited a PC with mystery stutter, clearing any manual useplatformclock flag and returning to automatic selection is a legitimate, safe step. Outside that, leave it alone and put your effort into drivers, thermals, and frame pacing, which move real-world smoothness far more.
Above all, measure rather than believe. Run an identical benchmark before and after any timer change and compare 1% lows. If the difference is inside normal run-to-run variance, the setting is a placebo on your hardware and you should move on to changes that actually register.
FAQ
Does disabling HPET give free FPS?
Rarely, and only on systems where a timer flag was misconfigured. Any gain is usually in the low single digits of 1% lows, not big average FPS jumps. Treat bold claims skeptically and measure on your own rig.
Is it risky to change?
No — it is fully reversible in BIOS or with a BCDEdit command. Just change one variable at a time and reboot so you can attribute any difference correctly.
Bottom line: HPET is mostly a myth-magnet. The genuinely useful action is ensuring no forced useplatformclock flag is set so Windows can auto-select the best timer. Test with a repeatable benchmark and LatencyMon; if nothing changes, leave it alone and spend your effort on drivers and frame pacing instead.
