Help get this topic noticed by sharing it on Twitter, Facebook, or email.

MPC-X: CPU load stays high untill double pressing STOP

I've bought me an MPC-X (2.2.1) to use for live performance, bringing progressive/techno.

On my first try I ran into the low RAM limiting the length of my set when using many different long multi-sample keygroup programs. I've overcome that part by using very short samples and chip tunes to merely use the sampler as a synth.

However, now on round 2, I run into high CPU usage after running just a few (effect heavy) sequences. And it stays high, even after pressing STOP once. Only after double pressing STOP, CPU drops. Funny thing it though, that running a single sequence only requires up to 20% CPU. So these lingering processes (I don't know what they are) - that don't even produce sound anymore - keep the CPU running high and I can't kill them without killing the live performance and double press STOP. That's a buzzkill.

Is there a way to stop these lingering processes and bring down CPU without double pressing STOP?

This problem may be resolved in a software update, a patch that improves the stability by identifying processes that are of no effect (silent/inactive/out of scope) and killing them.
3 people have
this problem
+1
Reply
  • Ok, that's interesting. We've never heard that before, but do know how to cope with
    it on software DAWs. That might not not help very much with "lingering processes" ,
    but reducing voices on synth playback, putting down decay/delay times in the effects
    and do a bounce to track on heavy load, does save cpu power. But that's all part of
    when producing in the studio, playing live is a quite different story. It would be quite
    good to know what's going on.

    PSOUND
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • Hey Siq,

    Thanks for posting!

    Sorry you're having some trouble with that! If you're using the MPC X effectively as a synth with insert effects and seeing that processes continue after pressing stop, it sounds like the MPC is still processing effect data on the Keygroup tracks.

    This behavior does sound like it is relative to the effects you've applied to your tracks as doing so will affect CPU usage. I'd like to get you in touch with our technical support department so we can work with you directly and take a closer look. In the meantime, have a look at this knowledge base article. It will help you optimize your MPC's performance:

    Akai Pro MPC X and MPC Live - System Resources Management

    Please look out for my email so we can assist you further - thanks!
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • Thanks for the reactions

    I'm aware of ways to lower the CPU and I will (have to) apply them. However, with a live perforance it's a matter of time before I run into the same issue. The issue seems that programs with a timed effect (such as reverbs & delays) will continue to require CPU even after they are no longer processing audible data, e.g. when they're practically silent. So even when the delayed sound that was used once, 10 sequences ago, can no longer be heard for 20 minutes .. it's program still required CPU.

    Anyway, I'm in contact with the tech team to discuss the issue. I'll keep this thread updated.

    - Klaas
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • FYI I am fairly certain your assessment is correct. Any effect that has an ongoing effect (eg delay or reverb) will continue using CPU indefinitely once active as they feedback into themselves. Theoretically they could be set to terminate when the input is inactive and the feedback hits 0 (which may or may not happen depending on accuracy and rounding) but the simplest approach is to leave them going.

    When you hit stop on your sequence the notes which are sounding all "key up" but, to avoid an abrupt termination of ambience, the effects in question continue to sound to let the ambience fade out naturally. When you hit stop a second time the play head is reset to 0 which means the ongoing ambience from the effects is no longer consistent with the current sequence time. This resets all envelopes and effects.

    Basically, this is "intended behaviour" to avoid, among other things, treading on your reverb tails. The question here is, how costly is it to start and restart the effect? If it is cheap, then I think the sensible solution would be to add a "noise floor" parameter to those effects which live beyond the current audio frame such that, if the amplitude of the feedback channel drops below that floor, the effect would terminate. This would mean that if you had an effect on a sparse channel it would only activate intermittently during playback (possibly allowing more effects in a sequence (as long as less are simultaneously active). If it is more expensive, the same process would work but it would only be termination eligible when the sequencer has stopped. Depending on the design and parameter range the floor level could be 0 (ideal) and therefore not even need new UI elements.
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • Hi Daniel,


    I could not agree more with your description of the issue and the possible solution and I hope that this issue is regarded seriously enough to investigate. The addition of the noise floor and detection of the post-effect signal below it could check if the process could be killed.


    Though I think adding the condition that no audio is playing (albeit silence) might make it easier for the effect to reinitialize/start again for it may then react on note-on, in stead of detecting activity on the signal - but that depends on how the effects trigger currently. I haven't tested that though. If effect already trigger (start using CPU) only when non-silent audio id playing, noise-floor detection is already present as input gate.


    That reminds me .. in recording mode, noise-floor detection/gating is already used. So technically code can be re-used :) Ok, maybe I'm getting ahead of it all, but I'm quite possitive about a possible solution.
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • yeah, i also encountered this. guess it's because of tape delays i'm using, they continue to self-oscillate, even if i don't hear them at all... and as a result, cpu meter goes to its limits after several sequences.
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated

  • *bump*

    Well, tech support has only given me an explanation of how this works, but no solution. I'd say this is a serious CPU leak, something that can and should be fixed in the software. It would be bad to advice to double-press STOP during a live act, just to flush the CPU load, I guess.

    So Nick, any updates no this?
  • (some HTML allowed)
    How does this make you feel?
    Add Image
    I'm

    e.g. kidding, amused, unsure, silly indifferent, undecided, unconcerned happy, confident, thankful, excited sad, anxious, confused, frustrated