Ember
All news

Ember 0.2: we measured FPS properly

For weeks people kept asking how much FPS Ember actually gains you, so we built a benchmark anyone can repeat: a world with a fixed seed, locked at noon, no rain, the camera in the exact same spot, and 30 seconds of measurement after a warm-up.

On 1.8.9, stock Minecraft averages 119 FPS (it ships with VSync and a 120 cap). With Ember it lands around 2540. On 1.7.10 we went from 118 to roughly 1890.

Where the difference comes from

  • Cached HUD. The HUD is drawn to a texture 60 times a second, and every other frame just pastes that image. That alone was worth about 50% in our tests.
  • No glGetError checks. Minecraft asks the graphics card whether anything went wrong several times per frame, and every one of those questions stalls the CPU until the GPU catches up.
  • A graphics profile built for PvP. No VSync or cap, VBOs on, no mipmaps, no entity shadows, no clouds. The launcher applies it once, on first launch.
  • Less stuff far away. Distant entities, chests and signs stop being drawn.

What we won't promise you

On a packed server with dozens of players on screen, the relative gain is smaller than in an empty world — the CPU and the entity count are in charge there. You'll still feel it, especially in the 1% lows, which is what you actually notice as stutter.

If you want to know exactly how the test works, the full method is on our Features page.