Knema – Frame Continuity Engine

Hello everyone,

I’d like to share an experimental project called Knema and open it up for technical discussion and feedback.

Knema is not an FPS booster, not an upscaler, and not a rendering feature.
Its focus is frametime continuity and stability improving perceived smoothness by managing how frames arrive, not by chasing higher benchmark numbers.

Philosophy

Most performance tools aim to maximize FPS. Knema takes a different approach:

  • Prioritizes frametime shape and consistency over peak FPS

  • Treats sudden spikes, oscillations, and pacing instability as first-class problems

  • Makes decisions based on real hardware limits, not ideal targets

  • Avoids aggressive boosting when the GPU is already at its physical ceiling

In short:

Stable, predictable frames are often more important than higher FPS.

What Knema Does

  • Monitors frametime distribution (mean, p95, jitter)

  • Detects GPU-bound vs engine-wait vs telemetry anomalies

  • Applies adaptive FPS and power behavior only when it improves continuity

  • Backs off automatically when limits are reached instead of forcing corrections

  • Handles common issues such as focus changes, telemetry gaps, and false utilization drops

Current State

  • Tested on multiple NVIDIA GPUs and several real-world gaming scenarios

  • Early results show reduced micro-stutter and more consistent frame pacing

  • However, testing is not exhaustive, and behavior may vary by system, driver, or game

Important Disclaimer

Knema is experimental software.

  • Not all edge cases have been tested

  • No guarantees are made regarding stability, performance gains, or compatibility

  • Use is entirely at your own risk

  • The author assumes no responsibility for crashes, data loss, hardware issues, or unintended side effects

If you choose to try it, please do so with caution and avoid critical or sensitive environments.

Download

Knema is available here:
https://sourceforge.net/projects/knema/

Feedback, criticism, and technical discussion are welcome.
I’m especially interested in perspectives on frametime analysis, pacing strategies, and GPU-bound decision logic.

Thanks for reading.