Problem with Diva and Ace in Mainstage

[Coustard]

M1 Max Macbook Pro, 64GB ram (plenty...), MacOS 12.5.1, Mainstage 3.6.1. Ace and Diva build 12092 MacARM64. So, everything up to date, I think. Motu Ultralite Mk3 Hybrid on USB.

Ordinary usage scarcely troubles the Macbook (CPU rarely above 15% as reported by Mainstage) but while the "patch remain" function (ie held notes persisting over patch changes) usually works reliably, if the new patch contains either Ace or Diva the outgoing audio is muted, the CPU meter goes to 100% and Mainstage presents a System Overload mesage, often reporting something like:

The audio engine was not able to process all required data in time. [channel strip name] > Diva used 120% of the available time.

I have "scan on startup" disabled, "Multicore" disabled, Accuracy set to fast - changing these parameters has no effect.

The only other 3rd party instrument that I use (Pianoteq) seems okay.

This is a long-standing issue for me using Intel Mac Minis on stage (with the Ultralite connected by FireWire) and I just assumed that they didn't have enough grunt but it persists on a new computer which surely has more than enough power.

Honestly, it's not a gig-killer, and I would not be without these instruments, but any wisdom would be much appreciated.

Thanks,

Coustard

[tasmaniandevil]

If Diva is using a CPU heavy preset, i.e. with many voices and long release times, this can put quite some load onto a single core of the CPU (even if it's an M1), if there is no multi-threading that distributes the voices over the available cores.

If you switch patches in Mainstage while playing them, Mainstage has to simultaneously load and play multiple patches. If the plugins of the new patch get loaded onto a core which already processes the old patch, it might be too much for a single core to handle.

There might also be a short spike in CPU usage at the moment a new patch is being loaded, as Mainstage needs to load all the settings and plugins. This spike might be too short to register in the CPU meter, but it could lead to an overload that disrupts the audio.

I did some quick tests loading multiple patches in Mainstage while keeping notes held, and so far haven't gotten any problems with CPU overloads on an M1 2020 MacBook Pro. But those were simple patches.

It could also be an issue connected to your audio interface and/or the audio settings, like sample rate and buffer size.

If you wish, we can try to help a bit in investigating the issue.
Best would be to contact our support directly, this way you could then also send us some example patches, so we can try to reproduce this more accurately.

https://u-he.com/support/#contact

[Coustard]

Thanks for that. It's not a question of Diva or Ace notes being held over a patch change (which I haven't checked), it's specifically (as you alluded to in your paragraph 3) that when holding notes on a Mainstage internal instrument (Sampler, say) and changing to a patch that contains Diva or Ace causes the CPU spike and notes to be cut off.

It's a pretty large concert - I typically have 20+ patches, many of them containing 20+ instruments - so I'll start with something smaller and see if I can find exactly what causes it.

Thanks for your help

[Coustard]

Okay, solved. Sort of.

I have buttons on my midi controllers set within Mainstage to send All Notes Off and All Sound Off to the destinations of my various keyboards. Getting rid of them (or some of them) reduces the CPU usage on patch change enough that incoming patches with Ace or Diva in them will no longer kill all sound. I've run out of time to investigate further but just in case this sheds some light or helps somebody else...