Docs/Using Frametide

Testing in your game

A test answers one question: which setup runs your game best on your PC? The app switches presets for you; you play the same scene each time while it measures every frame.

How it measures#

Frametide records the time between frames with PresentMon, Intel's open-source tool that reads frame events straight from Windows. Nothing is injected into the game and its memory isn't touched, so it works with anti-cheat. It also samples CPU and GPU load.

Set up a test#

1. Start your game first, then open Test in game and pick it under 1. Start your game, then pick it here. The list shows running programs and updates every time you open it. Pick the game itself, not its launcher.
2. 2. Which presets should be compared? Tick the presets. Your original settings is ticked and required on the first test of a game, because everything is compared against it. Later tests reuse those runs; tick it again after a big driver or game update.
3. 3. How long?
   • Quick check: 3 minutes, 1 run per preset.
   • Recommended: 5 minutes, 2 runs per preset.
   • Thorough: 10 minutes, 2 runs per preset.
   • Custom: set minutes per run, runs per preset and seconds to get ready under More options.
   The line below shows the total, e.g. “1 preset × 2 runs = about 11 min of playing”. With 2 or more runs per preset the app can tell real gains from random ups and downs.
4. Click Start test.

During the test#

For each step, the app applies the next preset and shows what to do:

1. Go to the place in the game you always test in.
2. Press F9 in the game. Measuring starts 5 seconds later, a sound marks the start, and another sound marks the end. You don't have to leave the game.
   Or click Start measuring in the app and switch back to the game within the seconds to get ready (30 by default; that time isn't counted).
3. Play normally until the end sound. The app shows the result of the run and prepares the next preset.

Presets alternate (A, B, then B, A), so heat and background activity hit every preset equally. Skip this preset moves on; Stop test ends the test and switches back to your original settings. Runs measured so far are kept.

Presets that need a restart#

Some settings only take effect after a Windows restart. When a preset needs one, the app shows Restart Windows now. After the restart and signing in, Frametide starts by itself and continues the test where it stopped.

Continue without restart (or turning off Restart Windows when a preset needs it under More options) is faster but less accurate: settings that need a restart aren't active yet, and those runs are marked with ⚠ in Results.

Getting reliable numbers#

The more alike your runs are, the smaller the differences Frametide can detect. If your runs vary by 10%, a preset that's 5% faster can't be told apart from luck.

• Same scene, same way. A fixed route, the same map, the same replay or the game's own benchmark. A practice range or a bot match works well in shooters; avoid scenes that depend on other players.
• Start measuring after loading. Let the level finish loading and the first shader compiling settle (the first minute after a driver or game update is often stuttery).
• Use F9 instead of alt-tabbing: switching windows itself can cause hitches.
• Close what you don't need: browser tabs with video, launchers downloading updates, recording software (unless you always run it).
• 2 runs per preset at least. A run that went wrong can be left out later in Results.
• Same game settings for every run: resolution, quality, frame limit. A frame cap in the game hides any difference in average FPS (the lows still count).