CPU management

The PolyPulse offers a great variety of sequencing and sound design possibilities, however the processing power of the PolyPulse is not unlimited. Some care has to be taken to not overload the CPU as CPU usage above 100% can result in glitches or crashing the firmware. This page is meant to give insight into how CPU usage might be reduced to get the most out of the PolyPulse.

Some general facts about CPU usage:

  • By far most CPU is used by sound engine voices and audio effects.

  • The base CPU usage is around 20%. This amount can’t be lowered by changing any of the settings.

  • Sequencing has almost no effect on CPU usage: triggering X amount of voices very fast will result in roughly the same CPU usage when triggering the same amount of voices slowly.

  • Peak CPU usage above 100% will result in underruns (short bits of silence and clicks) and might crash the firmware if CPU usage stays above 100% for long amounts of time.

  • The CPU usage meters are updated every 250 ms and show the average usage.

  • As the CPU usage rises above 70% the PolyPulse display will refresh less often.

CPU monitor

The CPU monitor can be used to gain insight in how CPU is consumed in a running project.

Press edit+settings and select CPU monitor to open the CPU monitor.

Press yes to exit the CPU monitor.

../_images/cpu-monitor.png

In the CPU monitor the individual CPU usage of all 5 tracks and audio effects is shown. Additionally it shows the CPU used by the output FX chain, the number of underruns that have occurred since the firmware started and the total CPU usage.

For each track it shows from left to right:

  • sound engine type

  • amount of voices playing

  • the CPU usage per voice

  • total CPU usage of that sound engine

  • audio effect type

  • CPU usage of audio effect

Average CPU usage measurements

To give an idea how the sound engines and audio effects relate to eachother in terms of CPU usage, the following measurements have been taken using the CPU monitor.

../_images/cpu-sound.svg
../_images/cpu-sound-dark.svg
../_images/cpu-afx.svg
../_images/cpu-afx-dark.svg

Sound engines

Shown below is the expected average CPU usage per voice when continously playing 8 voices. The actual CPU usage might be slightly higher if less voices are playing or if modulators are assigned to sound engine parameters.

Sound engine

Mode

% per voice

Notes

Additive

1 partial

5.6%

For every 4 partials not used the CPU usage is reduced.

16 partials

6.2%

32 partials

6.8%

Resonator

noise

5.6%

noise + samples

6.2%

FM

1 operator

4.1%

Set the volume of unused operators to -inf to reduce CPU usage.

2 operators

4.9%

3 operators

5.7%

4 operators

6.4%

Subtractive

filter: 12dB

3.3%

filter: 24dB

4.9%

Granular

short grains

2.7%

The grain envelope affects the CPU usage as it determines how often a new grain has to be picked from the sample list.

long grains

2.0%

Quad engine

kick

1.9%

snare

3.3%

ratchet

1.7%

metal

2.3%

sample

1.7%

2xsample

2.8%

sawstack

2.3%

FM

2.6%

External

0.1%

Audio effects

Shown below is the expected average CPU usage per audio effect when continously processing 2 channels. The actual CPU usage might be higher if more channels are processed or if modulators are assigned to parameters.

Audio effect

Mode

% per FX

Notes

Reverb

clean

9.0%

CPU usage stays roughly the same even when processing more than two channels.

lossy

9.0%

noisy

9.0%

14-bit

12.0%

12-bit

12.0%

Fade delay

8.0%

Limiter

7.0%

Compressor

6.2%

Delay

clean

3.4%

digital

4.6%

analog

5.3%

bucket

4.1%

Multiband

3.2%

FX placed after multiband will process more channels and therefore use more CPU than listed in this table.

Equalizer

2.5%

Drive

soft

2.2%

medium

1.7%

hard

1.7%

fold

2.1%

Filter

1.7%

Amp mod

1.7%

Patch

1.2%

Reroute

1.2%

Stereo

1.0%

Mono

0.3%

Note

The expected average CPU usage measurements listed on this page are made with firmware version 0.2.1. No rights can be derived from these measurements.