Should you cap your FPS? Frame limiters explained

Your game runs at 300 frames per second on a 144 Hz monitor. One guide says cap it; another says an uncapped game has the lowest latency. Both can be right, because a cap changes several things at once, and which of them you care about decides the answer.

What a cap actually does

A frame limiter holds the game to a set rate by making it wait. Below the limit it does nothing: Blur Busters' G-SYNC 101 closing FAQ says any limiter, in-game or external, "is only active when the framerate is sustained above the set FPS limit". While it is active, three things change.

• The frame time steadies and the queue empties. With a limit the machine can sustain, Blur Busters writes, "a constant framerate/frametime target is now ensured", and the pre-rendered frames queue "effectively becomes '0'". That queue is where the CPU's finished work waits for a busy GPU. NVIDIA's Reflex article says it keeps the GPU fed "but can introduce latency", and that "Latency is typically higher in GPU-bound cases because of the render queue". A good in-game limiter, in NVIDIA's words, will "stall the game in the right spots".
• A VRR display stays in its range. Above its maximum refresh, G-SYNC and FreeSync stop working; the G-SYNC, FreeSync and V-Sync guide explains what happens there.
• The GPU stops running flat out. NVIDIA, describing its driver's power management: when the GPU "is not saturated with work there is an opportunity to save power by reducing GPU clocks". The Prefer Maximum Performance mode overrides that and keeps clocks high, NVIDIA adds. Whether a cap lowers temperatures and fan noise on your card is something to check on its sensors, not assume.

What it costs is the frames above the limit. And if the CPU is the limit already, there is little queue to remove: NVIDIA notes that "being CPU bound is a lower latency state than being GPU bound".

Where the limiter sits

Blur Busters measured limiters against each other:

So use the game's own limiter where there is one, and an external one where there is not. Even an external cap can beat no cap: within the G-SYNC range, Blur Busters explains, a frame rate held by RTSS "can actually have 1 frame less input lag than the same uncapped framerate" with the pre-rendered frames setting at 1, because the held cap empties that queue. For plain V-Sync without VRR, Blur Busters prefers RTSS for its steadier frame times.

Reflex and Anti-Lag sit between the two ideas. NVIDIA calls Reflex a "dynamic" framerate limiter: it times the game's work so the render queue stays empty, without locking you to one number. AMD's Anti-Lag 2 SDK documentation describes the original Anti-Lag as a delay in the driver that keeps the CPU from running "too far ahead" of the GPU; Anti-Lag 2 moves that delay into the game, "just before the user controls (mouse/gamepad/keyboard) are sampled", which is why each game has to integrate it. Valve added Anti-Lag 2 support to CS2 in May 2024.

Setting a cap

CS2. The Frame Pacing section of the Advanced Video settings has Maximum FPS In Game and Maximum FPS In Menus, which, per Valve's June 2024 release notes, "control the fps_max and fps_max_ui convars respectively". Valve's August 2025 update notes say "fps_max can no longer be changed while connected to a server", so set it before you join. With G-Sync, V-Sync and Reflex all on, CS2 limits itself to "your display's refresh rate or slightly lower".

NVIDIA. Blur Busters' steps for Max Frame Rate: under Graphics in the NVIDIA app or Manage 3D settings in the classic Control Panel, set it to On, choose the limit and apply. One exception, as of October 2026: NVIDIA's March 2026 announcement still says the Dynamic mode of DLSS 4.5 Multi Frame Generation "is currently not compatible with frame rate limiters and V-Sync".

AMD. AMD's advice for FreeSync users fits in one line: when the frame rate regularly exceeds the refresh rate, "turn on V-Sync or cap your framerate". For the settings inside AMD Software, see AMD Software settings for games.

Choosing the number

With G-SYNC or a G-SYNC Compatible monitor: at least 3 below the maximum refresh (57 at 60 Hz, 97 at 100 Hz, 117 at 120 Hz, 141 at 144 Hz), with V-Sync on in the driver. Blur Busters found 2 below was enough in its tests and recommends 3 because limiters differ in accuracy; going lower than needed "can actually slightly increase input lag". In a game with Reflex, with VRR and V-Sync on, Reflex picks the limit for you. With FreeSync on an AMD card, AMD's own advice is the one above: V-Sync or a cap.

Without VRR, V-Sync off: the cap is not tied to the refresh rate, so its job is to keep the GPU from saturating. Set it where your machine holds it in the busiest scenes of your game, not the quiet ones: a cap the frame rate keeps falling under is inactive exactly when it matters. How this compares on latency with VRR is one of the points where sources disagree; G-SYNC, FreeSync and V-Sync sets out both sides.

Without VRR, V-Sync on: the cap does not remove V-Sync's own delay, which Blur Busters measured stacking with the limiter's, and Blur Busters suggests RTSS here for its steadier frame times.

Check it with a frame-time graph

A held cap shows as a frame-time line at about 1000 divided by the cap: 7.1 ms at 141, 4.2 ms at 237. Expect it flatter with an external limiter than with an in-game one, which Blur Busters says targets an average and lets single frame times vary. Spikes above the line are frames the cap could not hold. A line that sits above the cap's value all the time means the limiter is never active.

• PresentMon traces "CPU, GPU, and Display frame durations and latencies", and its capture application draws real-time graphs and saves per-frame CSV files; how to show an FPS counter covers it.
• NVIDIA FrameView measures "frame rates, frame times, power, and performance-per-watt on a wide range of graphics cards", using PresentMon underneath. Its power reading shows whether the cap changed what your card draws.
• w0a's Benchmark page measures frame times while you play your own game; reading the benchmark explains how to compare two runs honestly.

Measure before and after: the same scene, several captures each way, and compare the 1% lows rather than the average (high average FPS but bad 1% lows). A frame-time graph shows pacing, not input latency: the latency figures on this page are Blur Busters' and NVIDIA's, not something your capture confirms, and the input lag no software can fix shows where the rest of the delay sits.

What does not help

• A cap far below what you need. Blur Busters' tests showed no real improvement past 2 below the refresh rate, and it notes that limiting lower than needed can slightly increase input lag.
• A cap as a stutter fix. A limiter only removes frames above the limit; a frame that arrives late stays late.
• Trusting the counter alone. An FPS counter sitting exactly on the cap says the limiter is working, not that the frame times under it are even.

Related

• G-SYNC, FreeSync and V-Sync: how they fit together
• High average FPS but bad 1% lows
• The input lag no software can fix
• How to show an FPS counter in any game
• CS2 low FPS and stutter: a measured checklist