The No.1 Website for Pro Audio
M1 Apple Silicon Experiences in Audio
Old 18th November 2020
  #1
Lives for gear
 
3 Reviews written
🎧 15 years
M1 Apple Silicon Experiences in Audio

A thread for discussing user experience running M1 AS Macs -

***** UPDATE 11/18/20 ***************
I found an error in the Logic Setup that was limiting my M1 to 4 Cores (Preferences - Audio - {Processing Threads was on "Automatic" changed it to 8 cores). (this test ad errors) Now 44.1k test is More than 128 Tracks!!!!

********UPDATE 11/18/20 #2 **************
Still found one more error in previous test (my bad)
Get 90 tracks at 44.1kHz for any buffer from 32 - 256 (see page 4)

1000 apologies for all the changes but I found that I still ran the previous test for multicore wrong (bunch of tracks at the top were muted - my Bad...)

So here is the Multicore test results for different buffers which actually made no real difference! The CPU core meters would flatline at the top first then let me add 4 to 6 more tracks before stopping from overload. No glitches ever.

256 44.1
Full meters 86 tracks
90 tracks to OL

128 44.1
Full meters 86 tracks
90 tracks to OL

64 44.1
Full meters 84 tracks
90 tracks to OL

32 44.1
Full meters 82 tracks
90 tracks to OL

**************************************
Original post left up for context...


New Logic Pro Benchmark on the $999 MBA AS
(bought as a simple office and remote rec machine)
CPU shows up as 4 Cores in Logics performance meter - Does show as 8 cores in Activity Monitor though
44.1 128 buffer
Does 60 tracks.

96kHz 256 Buffer
~25 tracks (Only tested for a few minutes)

Runs just slightly warm in ~20% of backside and around function keys
Memory ~ 6GB used

FWIW
I just ran the same test on my iMac 2020 i7 - CORRECTED ~ 120 Tracks at 44.1 Fans a blazing - and 44 at 96kHz - 95degC 2600 ish rpm fans.

I wouldn't say the MBA is a powerhouse - but it is the most basic machine and is impressive at more than 1/2 the performance on this test with zero fan noise and minimal heating. Will be very excited to see the higher end products next year!

PS: Will have to wait for Apogee to get either a Beta or V1 Apogee driver before I can do any real-world testing in my system...

PSPS - This Benchmark test
Anyone interested in my new benchmark to replace the dated Evan's test?
Attached Thumbnails
M1 Apple Silicon Experiences in Audio-screen-shot-2020-11-17-2.53.11-pm.jpg   M1 Apple Silicon Experiences in Audio-screen-shot-2020-11-17-2.56.45-pm.jpg  

Last edited by ProPower; 20th November 2020 at 08:58 PM..
Old 18th November 2020
  #2
Lives for gear
 
🎧 5 years
Looks like a great result for a passiv cooled device.

I'm wondering why it does a lot less track that the intel, when they they it's faster than an intel cpu. Let's asume it 8 cores vs 4 core. When you double the Apple numbers it's a bit faster, but not something like the 2-3 times that you read a lot. Which i7 is in your imac?

Was the buffer size and sound device the same for both?
Old 18th November 2020
  #3
Deleted 94711a4
Guest
Thanks for doing the test.

I was just wondering about one thing. Did you have every other app closed and was Logic the only app running when doing your test? It looks like there is some other stuff going on in your activity monitor window.
Old 18th November 2020
  #4
Lives for gear
 
signalpudding's Avatar
 
I must be missing something, but tracks of what??
Old 18th November 2020
  #5
Lives for gear
 
🎧 15 years
Fwiw, I'm getting 89 tracks on my 2010 MP (2x2.66GHz, High Sierra, LPX 10.4.8). This is just now, with some other programs opened (should be able to squeeze 1-2 more tracks out of it with everything but Logic closed).

And that's why I'm wondering already, because given all the other available benchmarks so far (Cinebench, Geekbench), the M1 (even in the MBA) should do better than my machine.
But then, it's quite strange there's 4 idling cores. I actually would've expected them to be used for playback tracks inside Logic. Also, apparently they're used in the mentioned benchmarks - otherwise the M1 wouldn't score better than my MP in those tests. Usually, those benchmarks transfer quite well to Logic benchmarks (get 30% better results there and you'll as well be able to run 30% more tracks in Logic, at least almost), but in this case they don't.

Also, I'd really be curious about singlecore performance in Logic - maybe a new test would be a good idea. For my personal needs, I'm never making use of all the power of my current machine (I'm not into huge orchestral things or such), but I could very well do with quite some more singlecore juice, running complexed plugin chains in realtime. Given the mentioned benchmarks, any M1 machine should almost slaughter my MP (looks like 3 times the horsepower in Cinebench), so I'd probably be a more than happy camper in case that was true, even if the multithreading results remain the same.
Old 18th November 2020 | Show parent
  #6
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by signalpudding ➡️
I must be missing something, but tracks of what??
Of this Logic benchmark test:
Anyone interested in my new benchmark to replace the dated Evan's test?
Old 18th November 2020
  #7
Lives for gear
 
3 Reviews written
🎧 15 years
Thus post had many incorrect measurements ---- Deleting it

Last edited by ProPower; 20th November 2020 at 08:58 PM..
Old 18th November 2020
  #8
Lives for gear
 
Jim Rosebrook's Avatar
 
🎧 10 years
Quote:
Originally Posted by ProPower ➡️

New Logic Pro Benchmark on the $999 MBA AS
44.1 128 buffer
Does 60 tracks.

iMac 2020 i7 - ~80 to 100 tracks at 44.1
2018 i9 MBP 6-core, Same test at 44.1 and 128 Buffer...I'm getting 76 tracks

So higher end Intel MacBook Pros doing better than an M1 in Logic.

But... upcoming AS 16" MBP with an M1x or whatever should be a nice jump when they happen.
Old 18th November 2020
  #9
Lives for gear
 
3 Reviews written
🎧 15 years
************AGAIN ignore this post - had not corrected for using all 8 cores - sorry :-)

No real change from OS update. But I do see 4 cores doing all the Audio work while others are fairly idle...

Also lowest buffer at 44.1 is still 32 samples

I do find it very impressive to look at a $999 laptop that can do these 60 Tracks pretty much full CPU load and it is still silent and running just a little warm after 15 minutes. :-)
Attached Thumbnails
M1 Apple Silicon Experiences in Audio-screen-shot-2020-11-17-5.44.10-pm.jpg  

Last edited by ProPower; 20th November 2020 at 08:59 PM..
Old 18th November 2020
  #10
Lives for gear
 
🎧 15 years
I'm still wondering about the differences between the usual computer benchmarks and Logic performance. Something seems to be not right. Also, there's another video showing some FCP operation - with the M1 MBP beating a high end Intel model (I even think the i9 one). So why does that not apply to Logic?
Old 18th November 2020
  #11
Here for the gear
 
Very confused about those idle cores... could you fill us in on what plugins are being used on each track?
Old 18th November 2020 | Show parent
  #12
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by right_to_rage ➡️
Very confused about those idle cores... could you fill us in on what plugins are being used on each track?
Just download the benchmark test (link in one of my posts).

Edit: Here we go:
Anyone interested in my new benchmark to replace the dated Evan's test?
Old 18th November 2020
  #13
Gear Addict
 
🎧 15 years
The M1 has 4 high performance cores and 4 power efficient cores, so my guess is that only the high performance cores are working.
Old 18th November 2020 | Show parent
  #14
Lives for gear
 
Quote:
Originally Posted by mustgroove ➡️
The M1 has 4 high performance cores and 4 power efficient cores, so my guess is that only the high performance cores are working.
Yes, that would be my take on the matter - just as throttling and speed-step often causes enormous problems in DAW applications I don't see why allowing DAWs to run on these slow cores would be any different.
Old 18th November 2020
  #15
Lives for gear
 
Lady Gaia's Avatar
The power-efficient cores are generally used to schedule non-latency sensitive tasks. If Logic scheduled work on all available cores you'd wind up with unpredictable performance, which is a major problem for real-time tasks. Instead, they're able to soak up a lot of housekeeping tasks for the operating system so those don't interfere with your audio work.

These are entry-level machines and do seem pretty capable for what they are, but I expected we'll see an 8+4 core design in time that ups the ante considerably.
Old 18th November 2020 | Show parent
  #16
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Lady Gaia ➡️
The power-efficient cores are generally used to schedule non-latency sensitive tasks. If Logic scheduled work on all available cores you'd wind up with unpredictable performance, which is a major problem for real-time tasks.
But Logic is doing that already anyway. Every playback track is set to a higher buffersize already (1024 samples, IIRC).
Old 18th November 2020
  #17
Lives for gear
 
LASTLAVGH's Avatar
Check out this post and video from the Protools 2020 thread (and bear in mind this is using Rosetta even)

Quote:
Originally Posted by DigitalDrugLord ➡️
Did a "dirty" live with the M1 Air with 2020.11: https://www.facebook.com/504959982/v...8072415354983/

All I can say is I have over 100 computers.. nothing I have has ever performed this way.. this was broadcasted to my audio friends on FB, but thought you guys would like to see one of the M1 Macs on the latest PT


The benchmark of running a bunch of instances of Sculpture synth is definitely one metric but it may not tell the whole story.
Old 18th November 2020
  #18
Lives for gear
 
3 Reviews written
🎧 15 years
Did a render of 3.5 minute iMovie:
1080P
ProRes
Compressor = Faster

2020 iMac i7 5500 XT
45 sec

M1 MBA
60 sec
Old 18th November 2020 | Show parent
  #19
Gear Addict
 
🎧 5 years
Quote:
Originally Posted by LASTLAVGH ➡️
Check out this post and video from the Protools 2020 thread (and bear in mind this is using Rosetta even)


The benchmark of running a bunch of instances of Sculpture synth is definitely one metric but it may not tell the whole story.
“...Audio never skipped, audio never skipped!!...” LOL

Well, I’m officially intrigued. I reckon this Logic benchmark isn’t real life situation but I got surprised by this video. PT runs surprisingly well. Wasn’t expecting that.

I hope more real life mixing sessions videos appear showing what M1 can do in Logic, cos I was somewhat underwhelmed by these structure numbers. I wonder how many IK Tape 80 tracks it would be able to handle, for instance. That’s a plug I’d actually put on every track.. Or a session running Acustica channelstrips on every channel... I mean, plugins that actually make sense to be on every track.
Old 18th November 2020 | Show parent
  #20
Lives for gear
 
Quote:
Originally Posted by LASTLAVGH ➡️
Check out this post and video from the Protools 2020 thread (and bear in mind this is using Rosetta even)





The benchmark of running a bunch of instances of Sculpture synth is definitely one metric but it may not tell the whole story.
Well, given that Rosetta 2 executables seem to be 30-50% slower than native code (which is impressive for a post-compile translation across CPU architectures - genuine well done to Apple.) I can't see how this can be better than "a 100 PCs" the guy owns - unless they're from 2010!
Old 18th November 2020 | Show parent
  #21
Lives for gear
 
LASTLAVGH's Avatar
Quote:
Originally Posted by Scragend ➡️
Well, given that Rosetta 2 executables seem to be 30-50% slower than native code (which is impressive for a post-compile translation across CPU architectures - genuine well done to Apple.) I can't see how this can be better than "a 100 PCs" the guy owns - unless they're from 2010!
That guy is not kidding, if you follow him on some other threads you'll see he has an extremely high end collection of up to the minute gear.
He said it murders his 16" i9 MPB.
Old 18th November 2020
  #22
Gear Maniac
 
vocalnick's Avatar
 
🎧 15 years
Thanks for reporting experiences so far. I’m on the cusp of buying one to play with, but had a couple of questions.

Has anybody tried running x86 plugins via Rosetta? ie would something like NI Komplete work in the interim (likely with a performance hit) without needing to be specifically ported across?

EDIT: Just watched the video above - that looks promising!

Similarly, any chance at all that drivers might work via Rosetta? I expect this is particularly unlikely...
Old 18th November 2020
  #23
Gear Maniac
 
🎧 5 years
Anyone to test the m1 with a Dante converter running Dante software under rosetta 2 ?

Also interested to know if m1 works with native instrument under rosetta2

Thanks !
Old 18th November 2020
  #24
d3l
Gear Nut
 
🎧 5 years
So at the moment we know that Logic & PT both work. Some plugins work, but not necessarily all. So the million dollar question at the moment is, do the interfaces work and if they do, how?
Old 18th November 2020
  #25
Lives for gear
 
Lady Gaia's Avatar
Interfaces are going to be limited to class drivers until there are Universal versions of Big Sur drivers available, as this is one area where Rosetta translated binaries simply aren't supported. I expect the majority of the work is in migrating away from driver models that aren't supported in Big Sur, and in the QA process for the resulting driver. Compiling it for Apple Silicon is likely a pretty trivial part of the process.
Old 18th November 2020 | Show parent
  #26
Gear Maniac
 
Quote:
Originally Posted by ProPower ➡️
No real change from OS update. But I do see 4 cores doing all the Audio work while others are fairly idle...

Also lowest buffer at 44.1 is still 32 samples

I do find it very impressive to look at a $999 laptop that can do these 60 Tracks pretty much full CPU load and it is still silent and running just a little warm after 15 minutes. :-)
Just by looking at activity monitor it looks kind of not right.
Core 1-3 are idle, 4,5 are busy, 6 is idle and 7,8 are busy
That’s some strange allocation.
Since Pro Tools under Rosetta uses all 8 Cores of a MBA I would assume that an AS native Logic should do the same.
Old 18th November 2020 | Show parent
  #27
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Jolosch ➡️
Since Pro Tools under Rosetta uses all 8 Cores of a MBA I would assume that an AS native Logic should do the same.
Every bit this. There's something fishy going on with Logic.
Tough to believe, actually, given they should've had way more time to prepare Logic for the M1 launch.
Old 18th November 2020 | Show parent
  #28
Lives for gear
 
Quote:
Originally Posted by Lady Gaia ➡️
Interfaces are going to be limited to class drivers until there are Universal versions of Big Sur drivers available, as this is one area where Rosetta translated binaries simply aren't supported. I expect the majority of the work is in migrating away from driver models that aren't supported in Big Sur, and in the QA process for the resulting driver. Compiling it for Apple Silicon is likely a pretty trivial part of the process.
^ That approach makes complete sense to me.
Old 18th November 2020
  #29
Lives for gear
 
🎧 15 years
Something regarding the benchmark just crossed my mind: There's a "PtVerb" used on all these tracks. This has been rendered "legacy" with previous versions already, could it possibly be that these legacy plugins aren't optimized for silicon Macs and that proper multithreading won't work because of that? Yeah, I think it's quite unlikely, but if you don't mind, @ ProPower , you could possibly throw all these PtVerb instances out to see whether something will change.
Old 18th November 2020
  #30
Deleted b1c557d
Guest
Is it possible to install older versions of Logic Pro?

Might be interesting to see if that changes anything.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 4616 views: 610125
Avatar for smoke
smoke 1 week ago
replies: 97 views: 36686
Avatar for mowmow
mowmow 15th September 2010
replies: 835 views: 176276
Avatar for jlgrimes11
jlgrimes11 30th August 2020
replies: 518 views: 40781
Avatar for audslu
audslu 28th December 2015
Topic:
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump