Quick performance test: ARM vs x86 Rosetta on M1 Pro MacBook Pro

I was very curious to compare the native ARM and x86 versions under Rosetta on my M1 Pro MacBook Pro. This is not a real use case, but gives a little insight on performance improvements with the native version…

My test was to stack up as many Omnisphere instances playing a four note chord progression on a relatively complex patch. Running under Rosetta I was able to play 148 tracks/instances before the audio would break up with a buffer size set at 256 samples. With the same project loaded in the ARM version I was able to increase this to 212 tracks. And this is on a laptop! :exploding_head: Make of this what you will. I was blown away with both results! YMMV.

Now time to make some music :slight_smile:



@SteveHelstrip: thanks for the feedback! We’re happy to hear that!

And I suspect you’ll feel even more of a difference if you actually had the Intel computer on your lap! :fire: :sweat_smile:

I have an 8700K desktop here and it doesn’t come close! :wink:

Just confirm Steve, by Intel version, you mean, running Cubase 12 under Rosetta 2?

Yes, running under Rosetta. I’ll amend to x86 code under Rosetta.

Hey @SteveHelstrip , thanks for sharing your results!
Unfortunately, I am having some strange results:

The same test project with some 77 Omnisphere tracks playing a fairly complex stacked multi, runs better on Cubase 11 Pro, than on Cubase 12 running through Rosetta. And it runs almost the same on C11 and C12 in M1 native mode…
Those 77 tracks are a limit before glitches appear under 256 buffer size and 48kHz sample rate. And I added a Retrologue instance for testing live playing while all those tracks are playing.
I also added a Reverence and a Vintage Compressor to add a bit more load on the computer :slight_smile:
I am using a MacBook Pro 14" base model, so it’s 16GB of RAM.

Maybe I am missing something here, so I would like to ask you if you could share your test project here? Here is mine:



I made a new test project, consisting of Omnisphere instances and one Retrologue instance. No inserts or sends. The results are as follows:

C11: 175 Omnisphere tracks
C12 under Rosetta: 165 Omnisphere tracks
C12 Native ARM: 234 Omnisphere tracks

C12 performance under Rosetta is very disappointing, since most of us will still have to use it in order to use VSTs which still don’t support M1 (which is like all of them :smiley: )…

If I made some important omission, I would like to hear from all of you.



I did a test with a virtual instrument project and some plugins.

I found that C11 under Rosetta worked just slightly better than C12 under Rosetta (I dare say with an actual Intel computer that might yield a different result).

I found that C12 native mode was about 25% better, i.e. 25% less on the performance meter which is great.

using the Steinberg demo that has 70 tracks with all the plugins on it, I found that it worked about the same in C12 Rosetta or C12 native but that might be because it uses so little power anyway.

But everyone uses different combinations of plugins and instruments. We all have different Audio Devices at different buffers and use our DAWs for different things. So it might be impossible to get a performance difference that is not unique to way someone works.

I’d like to think the C12 Native performance increase is about 25%, but it might not be that on all projects and all stacks.

That’s about what I get. Cubase 12 rosetta is slightly worse than 11 and 12 native ia quite a bi better than rosseta.

1 Like

So the native M1 version is performing about 35% better than the Rosetta version.
Cubase 12 offers quite a performance bump for Mac users, impressive.


True, but only few plugins work in M1 native mode… Hopefully that will change soon :slight_smile:

1 Like

iLok is now compatible, I think that will be a huge boost.

1 Like

Great news, thanks for posting, very encouraging for those of us wondering whether to bite the bullet and go Apple silicon.


plus you can open M1 native VST2’s using Blue Cat Patchwork in C12 Native Apple Silicon. So that helps with the whole VST3 issue. Of course the VTS2 has to be native but most of UAD is already native, on the UAD website they state they are not native but on their forum they tell you they are native except a couple still to go, plus I can use them via patchwork in C12 native so they must be lol. A lot of instruments that are native are VST2 as well so I can use them. So Patchwork can be very useful.

Oh that’s interesting. Trying the demo now. You can also open AU plugins inside Cubase 12. This could be a game changer, if it works well.

1 Like

Yes I noticed that too. So my Roland native plugins that are only AU can be used in C12 native too.

It just does not use Rosetta on AU plugins like it would in logic or Ableton.