Clipping without
the FPS hit.

Two ways to clip, only one costs frames.

Clipping happens in one of two places. During the match, which is what a replay buffer or an instant-clip hotkey does: something is capturing and encoding your frames the whole time you play. Or after the match, on a file that is already sitting on your drive, with the game closed.

Only the first shape can cost you frames, because only the first shape is running while the game needs the hardware. CutBench is the second shape. It does not install into your game, it does not sit in the present path, it does not hold a buffer open, and unless you launch it yourself it is not running while you play. That is not a low-overhead mode or a clever setting. There is nothing there to cost you anything.

The honest flip side: you do not get the clip ten seconds after the play. You get a ranked set of vertical clips once the session is over and the processing pass has finished.

What in-game capture actually costs.

People blame "recording" as if it were one thing. The cost is really spread over at least six separate jobs, and which one bites depends entirely on your build and your settings.

• The encoder. On NVIDIA cards NVENC is a dedicated hardware block on the die, separate from the shader cores that render your game. That is why hardware encoding is the cheap path, and why the old folklore about recording halving your frame rate is mostly out of date. Cheap, though, is not free.
• The copy. Before anything can be encoded, each captured frame has to be pulled out of the render pipeline. That copy sits in the same present path your game uses to put a frame on screen, and it happens on every frame you record, not once per clip.
• The CPU side. Muxing, timestamps, audio mixing, writing the file, plus any overlay or hotkey hook injected into the game process. In a CPU-bound game, and competitive shooters at high refresh rates usually are, this is the part that shows up in your frame times.
• The disk. A rolling replay buffer writes continuously for the whole session. If those writes land on the same drive the game is streaming assets from, you get hitches that never move your average FPS number at all.
• VRAM. Capture buffers live in video memory. On a 6 GB or 8 GB card at 1440p with a long replay buffer, that is memory your game no longer has, and running short on VRAM does not cost a few percent, it costs stutter.
• Software encoding. If the recorder is set to x264 rather than the GPU encoder, the encode runs on your CPU cores at whatever preset you picked. This is the most common reason a recording setup wrecks a game, and it is a settings problem rather than a law of physics.

None of that makes in-game clippers bad software. It makes them software with a running cost, which is a normal trade and one plenty of players are happy to make.

Averages lie. Check your 1% lows.

Before you change anything, get a number for your own machine. Figures published by other people are close to useless here, because the answer depends on your CPU headroom, your resolution, your capture settings and the game.

• Use the same 60 seconds twice. A built-in benchmark, a practice range, or a fixed route on a quiet server. Anything with live players in it varies more than the effect you are trying to measure.
• Log frame times, not just FPS. Most GPU overlays and benchmark tools report 1% and 0.1% lows. Capture overhead usually shows up there first, as a rougher frame time line, long before the average moves at all.
• Change one thing per run. Encoder first, then resolution, then bitrate, then buffer length. Flip all four at once and you learn nothing.
• Write the recording to a different drive. If the numbers improve, your problem was disk contention rather than the encoder, and no amount of encoder tuning was ever going to fix it.
• Compare against a genuinely clean run. Recorder fully closed, not just idle in the tray. Overlays and hooks can stay loaded in the game even when capture is stopped.

Do this once and you will know whether you have a problem worth solving. Plenty of people find their overhead is small enough to ignore, and that is a perfectly good outcome. This page is not trying to talk you out of a setup that already works.

Same job, different moment.

Describes how the two approaches work by design, checked 14 August 2026. The overhead of any specific recorder depends on its settings and your hardware, so measure your own.

When an in-game clipper is the right answer.

Record with whatever you already use.

CutBench is not a recorder and has no interest in replacing yours. It reads the files your existing setup already produces, so everything you tuned over the years keeps working exactly as it does today.

• OBS Studio. Standard MP4 or MKV output. If you are chasing frames, point it at the NVIDIA hardware encoder rather than x264 and record to a second drive.
• NVIDIA ShadowPlay and the NVIDIA app. Manual recording or Instant Replay, both land as normal MP4 files.
• Medal. Its clips and its full session recordings are ordinary video files on disk, which is all CutBench needs.
• Outplayed. Same story. Point CutBench at the folder its captures already land in.
• Xbox Game Bar. Writes plain MP4 files into your Captures folder, which CutBench reads as they are.
• Steam game recording. Steam stores a background session as chunked segments rather than one finished video, so export the part you want to MP4 from Steam first, then point CutBench at that file.

There is nothing to install into your game, no overlay to load, no hook and no capture driver. CutBench never sees your game while it is running. It sees a file, later.

On settings: a clean 1080p60 recording is plenty of source for a vertical clip. Recording higher costs you more overhead during play and buys very little once the clip is cropped to 9:16 for a phone screen. Recorder defaults and output formats change, so the notes above were checked on 14 August 2026.

Play first. Clip after.

The obvious objection: does that not just move the GPU cost somewhere else? Yes, and that is the whole point. It moves it to a time when no game needs the GPU. Start it when you stop playing, or leave it running overnight, and the frames it uses are frames nothing else wanted.

More on the long-session workflow in how to clip a 3-hour stream, and on keeping the footage off the cloud entirely on the offline clip maker page. If you are still picking a tool, the best clipping software for PC puts the capture tools and the cutting tools in one table, and how an AI clip maker for gaming works covers what the ranking pass is actually reading.

Clipping and FPS FAQ.