Performance analysis of AVX2-enabled plugins (March 2021)
- Last Update
- 2021-03-14
This article reports performace measurement result related to AVX2-enabled plugins.
Test Environment
- Intel Core i7-6700K
- DDR4-2166 16GB main memory
- RME HDSPe AES
- Samplitude Pro X5
- Buffer Size: 256 samples
- A.O.M. Audio Plug-ins v1.11.4
Measurement
CPU Load[%] = Processing Time[second] ÷ (Buffer Size[samples] ÷ Sampling Frequnecy[Hz]) × 100
- Used time profiler-enabled private build. Other settings (like optimization) are same to release build.
- std::chrono is used for time measurement.
Result
Plugin Name | Settings | Generic Build CPU Load [%] | AVX2 Build CPU Load [%] | Load Reduction [%] |
---|---|---|---|---|
Cyclic Panner | LPF Link=ON | 0.21 | 0.21 | 0 |
Invisible Limiter | x16 oversampling | 7.02 | 5.74 | 19 |
Invisible Limiter G2 | Modern IV / Quality=#7 | 27.14 | 20.28 | 26 |
Nu Compresor | Defaults | 1.87 | 1.49 | 21 |
Sakura Dither | Defaults | 2.74 | 2.42 | 12 |
Stereo Imager D | Defaults | 0.64 | 0.57 | 11 |
tranQuilizr | Quality=5 | 9.12 | 6.03 | 34 |
tranQuilizr G2 | Extreme / Algorithm=II | 14.43 | 8.08 | 45 |
Wave Shredder | All blocks activated | 0.81 | 0.82 | -1 |