Yes, a little or a lot, depending on the recorder. In our July 2026 Cyberpunk 2077 test at 1080p on one PC with a GeForce RTX 5070, Flashback 0.6.22 cost 5.9% of average FPS; Medal cost 27.0%; NVIDIA ShadowPlay cost 28.4%.
Recorders that encode in software (x264 and similar) take CPU time the game would otherwise use, which shows up as lower 1% lows in CPU-heavy games. A hardware encoder is a separate part of the graphics card, so it usually costs the game much less. Check that your recorder is set to a hardware encoder.
In Flashback: Flashback uses the graphics card's hardware encoder through Windows Media Foundation.
Resolution, frame rate and bitrate all add encoding and disk work. 1440p or 120 FPS clips cost more than 1080p at 60 FPS, and you may not need them if clips end up on Discord or a phone screen.
In Flashback: Presets go from Competitive (1080p, 60 FPS, 10 Mbps) to High motion (1080p, 120 FPS, 24 Mbps). The default, Balanced, is 1080p, 60 FPS, 15 Mbps.
Game Bar's background recording, the NVIDIA app's instant replay, Discord and a third-party recorder can all be buffering the same game at once, each with its own encoder session. Pick one and turn the others' replay features off.
In Flashback: Flashback is one app with its own replay buffer, so it doesn't need the NVIDIA app or Game Bar recording alongside it. Turn their replay features off while it runs.
In-game overlays, automatic uploads and clip processing while you play all compete with the game for the same PC. Turn off overlays you don't use and let exports wait until after the match.
In Flashback: Flashback records the display instead of hooking into the game, and exports, trims and Discord copies run at below-normal priority so the game comes first.
Every PC, game and driver is different, so a number from someone else's setup is only a starting point. Run a game's built-in benchmark with the recorder off, then on with the settings you actually use, and compare average FPS and 1% lows.
In Flashback: That's how the published benchmark was run: the same settings in every recorder, against the same PC with no recorder running.
They come from one controlled run on one PC with a GeForce RTX 5070, measured July 23, 2026 against the same PC with no recorder running. Your result depends on the game, settings and hardware, and newer versions haven't been through the comparison yet. The full benchmark has the method, 1% lows, memory and raw data. Comparing recorders? See how Flashback compares with Medal and NVIDIA ShadowPlay.
