Reaktor 6: poor performance under Apple Silicon

"I am running my RME UCX-II @ 128 samples - and I hear zero latency for every instrument"

1. "You".
2. I'm using audio software monitoring. Latency is around twice as much (minus MIDI latency).
3. ETL of thw UCX at 128 samples is around 7.5ms. Latency of my M2 is 3.5 at 32. I want that, no matter how much you don't need it. It's the reason why I bought that very interface.

"Took me about 4 seconds to google this and found tons of hits on M2 drivers - in Mac AND in Windows

https://www.google.com/search?q=MOTO+M2+drive+rissues&ie=UTF-8

Or this specific example:

https://www.audiosciencereview.com/forum/index.php?threads/motu-m4-m2-driver-issues-system-process-high-cpu-motucoreuac-sys.43086/ "

I obviously googled as well, so these links are familiar. With the exception of Reaktor, I don't experience any such issues, not even remotely. Anything else is running smooth as butter. No idea why these folks are running into troubles.

See, when I'm saying explicitely (!) that I want to work at the lowest possible latencies in the opening post of this thread, why exactly do people try to educate me that it doesn't matter should I switch to 64 or 128 samples? And it's not that this was once but multiple times throughout this thread.

I have been investing a whole lot of time on whatever effects of latency since decades already. In fact, I have started "measure your interface latency" threads on both the international LUG (Logic User Group) and VST mailing lists back when, well, they were mailing lists (around the y2k). The guy who later started the low latency performance megathread on GS (Vince Curigliano) was part of one of these threads, AFAIR.

I did then proceed to do latency tests with colleagues and partially there's been quite some revealing results - sometimes proving what you'd think, sometimes hinting in the opposite direction (such as vocalists sometimes getting along better with a somewhat higher (!) monitoring latency - obviously for reasons). Etc. I could possible write an essay about these things.

Further, as I had been googling M2 experiences quite a bit before I bought it, I was aware of potential issues. Hence, the very first thing I tried was running a whole number of things under 32 samples (mainly involving Helix Native as I'm a guitar player in the first place). And everything I tried was working just fine. In fact, even incredibly well. Otherwise I would've sent the M2 back and had possibly bought a used original Babyface instead (V2 is a bit too expensive as a constant carry on, which is what I wanted). But as said, it was all fine - until I stumbled over Reaktor and Kontakt 7 Player. And while the latter appeared to be a messed up V10, so I could go back to V9, for Reaktor there's no such options.

To recap: Right now, the only (!) plugin causing massive problems at low buffersizes happens to be Reaktor. And it's no issue with the Intel version. So I completely fail to see how folks continue to tell me it's the (evil) Mac, my interface or my demand to work at 32 samples buffersize.

Tell me any plugin, I'll happily install it and give it a testride (obviously those Nebula things are out of the equation, they're still not exactly meant to be used on live threads). In case any of them shows the same issues as Reaktor, I'll happily take some things back (even if that would possibly just demonstrate that other plugins aren't exactly coded well for silicon, either).

"For your situation - probably time to discontinue using Reaktor and/or
Kontakt and simply carry on with what might work for you over there."

And that's precisely what I will do. I will keep using what I have (and what works) and look for alternatives and slowly say byebye to anything NI (unless they massively change their product support, which doesn't seem likely to happen, they're even saying so themselves).

"If you select Host or 44.1 fixed are the perfomance as poor?"

Every bit yes. "Host" is the default, so that's the setting when I noticed it was overloading pretty bad - but I also checked Fixed 44.1/48 kHz and some other settings (just out of interest). If at all, it's getting worse.

@Jeremy_NI, fwiw, the thing astonishing me the most would be that macOS activity monitor never shows any overloads. So, in laymen terms, it seems as if Reaktor can't utilize the available CPU overhead. Which is pretty weird to me.

Well, Reaktor is dead. NI even says so themselves (at least once you read between the lines).

I'm moving away from anything related.