Reaktor 6: poor performance under Apple Silicon
Comments
-
Ok, Super 8 was released in 2019. That's still not a dinosaur. U-He Zebra is almost 20 years old (cross platform since 2006) and still kicking a** big time (and obviously running incredibly well on this Macbook).
0 -
That behaviour displayed in your video may be due to e-cores. I doubt Reaktor standalone runs just on one core. Why should it? It may have multiple threads and OS may allocate one of them to e-core, then decide to assingn it to p-core, and instant later oh no, back to e-core….. And so on….
From what I read few years ago, there even were no means for application programmer to tell OS "I want this thread to run at this core. Or at least, I want to run it on e-core/p-core.." Maybe it has changed since then. But maybe not in Reaktor….
Why you do not try to restrict e-cores for test? Telling I am sure, Reaktor runs only on p-cores is proof, it is just belief/assumption.
And still no picture/video of individual cores' load…
There is a lot of problems with big-little also on Intels. But users may tame it by Lasso application. Big-little is rather new concept in PCs, so no wonder time critical applications fight with it. Some well, some not so well…
0 -
That behaviour displayed in your video may be due to e-cores. I doubt Reaktor standalone runs just on one core. Why should it?
Because it's a serial process that can't be broken down to parallel multitasks in realtime. As easy as that. This is DAW processing 101. And for the umpteenth time: No, it's got absolutely nothing to do with the E-cores.
0 -
And fwiw, on macOS' activity monitor, there's no overloads happening. It's just inside Reaktor (and Logic, but we can keep that aside as it's also happening in Reaktor/standalone).
So, it seems as if Reaktor simply can't deal with the CPU power available.
0 -
Fwiw, what I said in my last post is really, really suspicious. At the moment when Super 8 and Form are crackling like mad already while Reaktor is displaying "Over", macOS' activity monitor is never going above 40% (for a single core of course).
So it's pretty much as if Reaktor wouldn't be able to manage taking advance of the free CPU ressources.
Quite weird. And defenitely not exactly a prove for excellent programming…
1 -
Fwiw, not directly related but still related to the general way NI is working these days: I have just been checking out some libraries requiring Kontakt 7. Had to use the Player but that shouldn't matter.
Now, what could I possibly say? What a pathetic joke is this? I have not ONE single other plugin with such a horribly sluggish UI. In standalone mode, once you have and instrument loaded, when you enter preferences, clicking on the entries takes some seconds (!) for the click to register.
I absolutely can't believe that. I mean, it's their friggin' flagship product, if there ever was one. And it's performing as if I was on a computer from the mid nineties.
Absolutely not wondering about the Reaktor drama anymore. This company has completely lost it.
1 -
And fwiw: I installed an older version of Kontakt 7 Player (download location provided by NI, meanwhile business as usual: GDrive - apparently they can't even afford their own server space anymore…) and it's working 10 times better than the latest one.
I also remember reading about this elsewhere. Does anybody at NI test this stuff at all anymore? The sluggish performance of the latest version would've been considered a showstopper with a big red "do not release!" flag on it in any betatest I've ever been part of. Same with Reaktor, of course.
1 -
And yet another really strange thing (sorry it's again about Kontakt, but all this is more about general NI "tactics" anyway, so we can as well keep it in this thread):
When I initially installed the K7 Player, it installed version 10. Which was as ****** as it gets.
The alternate version I posted about above was 4.2. Which was working pretty fine (but not loading the "Hypha" library).
But obviously, Native Access now told me there was an update after finding V4.2. But that update was NOT V10 but V9, go figure! So I backed up V10 and V4.2 and performed the update. And guess what? While still a little slower than K6, at least once the libraries are loaded, everything is quite solid and fast.
So, bottomline: Apparently NI realized that V10 was rubbish and pulled it (NA doesn't show any update options anymore, now that I'm on V9). They're obviously not offering a "downgrade" option, which is bad.
But at least I feel confirmed now. I have been told that Macs are s***. I have been told my machine was under-spec'ed (muahaha). I have been told I was doing something wrong, I had no clue and what not (just as happened in this thread as well). Turns out neither was the case. At least K7P V10 was a t**d, yet it was released (at least for a while). Every bit the same with the latest Reaktor version (just that unfortunately there's no proper way to revert to an older version).
Edit: Whoops, I've just seen this is a known issue. So why don't they just notify their users?
0 -
@SaschaFranck Reaktor hasn't been updated for a while, that's a fact but there were no reports regarding perfomance of Reaktor 6 on Sillicon Macs so far. We would like to investigate, so could you tell us the following?
What kind of M3 processor is it? Max, Pro, Ultra? What is your operating system? Sonoma 14.4.1?
In what DAW is this happening? Is it also happening in standalone?
I know you really really want to have a 32 samples buffer size but we've seen cases where a higher buffer size would actually improve the performance, I know it sounds counter intuitive but it's worth trying, at least to troubleshoot.
I understand that it can be frustrating to get a brand new computer and experiencing lower performance than your old mac but please keep in mind that people out here are here to help. No need to be aggressive or use foul language. Let's try to stay respectful and constructive here.
These M3 macs are pretty new, the and M2 M3 Max, Pro and Ultra combined with the 14.4.1 Sonoma update brought a lot of issues, especially with Kontakt 7, as you can see here: Kontakt 7 window across multiple screens causing Kernel panic shutoff on Mac Studio! (KT-8785)
This is in no way pointing the finger on Apple, we gather as much info as we can to report to them and we check what we can do to mitigate the issue. This Reaktor 6 performance thing is new to us. If we need to investigate and/or file another report to Apple we'll definitely do it.
The Kontakt 7.9 / 7.10.1 was a totally different issue, there's a 7.10.2 update coming soon to definitely fix it for Sonoma users.
Also keep in mind it's a user forum and not a substitute for support. Of course we can continue the conversation here but if you're not at ease with responses from other users you can reach out to support here: https://bit.ly/NI_support_other Please keep it civil and respectful there as well. Thank you.
1 -
Hi Jeremy,
"What kind of M3 processor is it?"
"Plain" M3 (Macbook Air - they only come with that).
"Sonoma 14.4.1? "
Yes.
"In what DAW is this happening? Is it also happening in standalone?"
Logic, Garageband, Reaper, standalone (the video 1-2 pages back is made with the standalone version).
"I know you really really want to have a 32 samples buffer size but we've
seen cases where a higher buffer size would actually improve the
performance"I'm into this for long enough to know that larger buffer sizes improve performance. But I don't want to keep flipping my settings just to run 2-3 Reaktor ensembles. With anything else (and now even with K7P) it's set and forget. Has been like that since ages under Logic, one of the reasons I still quite like it.
"I understand that it can be frustrating to get a brand new computer and experiencing lower performance than your old mac"
The thing is, I'm experiencing fantastic performance. It's really just Reaktor completely crapping out.
"These M3 macs are pretty new, the and M2 M3 Max, Pro and Ultra combined
with the 14.4.1 Sonoma update brought a lot of issues, especially with
Kontakt 7, as you can see here: Kontakt 7 window across multiple screens causing Kernel panic shutoff on Mac Studio! (KT-8785)"That's all fine and dandy - but how comes I haven't experienced any issues with any other plugin so far? Not even with ones still running in Rosetta mode?
"Of course we can continue the conversation here but if you're not at ease with responses from other users"
I'm totally cool. But that doesn't mean I will silently accept being called clueless and what not. When in the end it's sometimes even just that people don't even bother to read properly.
"Please keep it civil and respectful there as well."
I'll try my best.
0 -
Because it's a serial process that can't be broken down to parallel multitasks in realtime. As easy as that. This is DAW processing 101. And for the umpteenth time: No, it's got absolutely nothing to do with the E-cores.
Of course it may be split to several threads…. I am a computer programmer, by the way….
Concerning CPU load…. Short spikes may not be visible in monitor. And threads jumping between e and p cores may introduce temporary latencies….
If you do not want to do the test with e-cores off, your decision. But, then you will not find if that matters or not….
0 -
"Of course it may be split to several threads"
Sure. But that would add latency.
"If you do not want to do the test with e-cores off, your decision."
I did. Everything the same.
0 -
No, it would not add latency if done properly. If nothing else interface of plugin/application may be placed to another thread. And in many cases also signal path. But it of course depends on type of work plugin does.
0 -
So, how comes no plugin is already doing it?
0 -
But that's just it. I am running my RME UCX-II @ 128 samples - and I hear zero latency for every instrument - not just Reaktor. I continue to be baffled on why 32 or even 64 samples are even a discussion point here.
And yes I do run Windows - but if this was a major issue with ANY VST in my collection I am certain I would have noticed something by now.
On that note - has anyone even stopped to investigate the MOTU audio drivers in this scenario?
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:
There could be something else afoot here besides just Reaktor OR assumptions on the M3 chip…
VP
1
Categories
- All Categories
- 19 Welcome
- 1.5K Hangout
- 62 NI News
- 785 Tech Talks
- 4.1K Native Access
- 16.5K Komplete
- 2K Komplete General
- 4.3K Komplete Kontrol
- 5.7K Kontakt
- 1.6K Reaktor
- 378 Battery 4
- 845 Guitar Rig & FX
- 429 Massive X & Synths
- 1.3K Other Software & Hardware
- 5.8K Maschine
- 7.3K Traktor
- 7.3K Traktor Software & Hardware
- Check out everything you can do
- Create an account
- See member benefits
- Answer questions
- Ask the community
- See product news
- Connect with creators