Hello, I did a few tests to help me decide whether I need a new computer and if so, which one.

14" M1pro - 10 core - 16GB ram - My current Computer
14" M3Max - 16 core - 48GB ram - Apple price in Germany 4570 Euro
14" M3Max - 14 core - 36GB ram - Street price In Germany 3570 Euro

Here in Germany you can get the M3Max for around 3570 euros and the M3Max can only be purchased directly from Apple for 4570 euros. So a difference of 1000 euros.

——————————————————————————————

Demo Song Live 11 Loop the part 49-57
44,1kHz, 128 Buffer Internal Soundcard

M3 Max 16 Core 185%
M3 Max 14 Core 170%
M3 Max 14 Core 157%
M1 Pro 10 Core 100%

M3 Max 16 Core 13x Demo Song (CPU Peak 81%) (377 tracks)
M3 Max 14 Core 12x Demo Song (CPU Peak 87%) (348 tracks)
M3 Max 14 Core 11x Demo Song (CPU Peak 80%) (319 tracks)
M1 Pro 10 Core 07x Demo Song (CPU Peak 80%) (203 tracks)

——————————————————————————————

YouTuber James Zahn did a test with the new M3pro. I recreate the same test with the M3Max.
Ableton Live 11.
44,1kHz, 1024 Buffer
Every track got a Audio Guitar wave + Neural DSP's Archetype: Nolly plugin

M3 Pro 11 Core 60 tracks 83% ( James Zahn test)
M1 Pro 10 Core 72 tracks 100%
M3 Max 14 Core 91 tracks 126%
M3 Max 16 Core 109 tracks 151%

——————————————————————————————

Alternately loaded a track of Serum and a track of Diva. Besides, there was still on every track
EQ Eight, ValhallaDelay, ValhallaVintageVerb, Compressor, Multiband Dynamics, Glue Compressor and Archetype Nolly Guitar-amp.

Serum and Diva were on the highest settings and played 6 notes at the same time.

Attached are the results.

44,1kHz
256 samples
Internal sound card

M1Pro 10 core: - 20 tracks
M3Max 14 core: - 36 tracks
M3Max 16 Core: - 38 tracks

That's impressive for the M3Max.

——————————————————————————————

Fan Noise @80% CPU after 15 min.

M1 Pro 10 Core - -40,6dB
M3 Max 14 Core - -40,5dB
M3 Max 16 Core - -35,1dB

The M3 MAX 16 core fan is approx. 5 dB louder!

Let me know if you want me to do any more test.


Location of the default Demo Song

Mac
Right-click on the Live application from the Applications folder in Finder and select "Show Package Contents".
Then go to this folder:
Contents/App Resources/Core Library/Lessons/Demo Songs

Windows
C:\ProgramData\Ableton\Live 11 Suite\Resources\Core Library\Lessons\Demo Songs


Some useful information about the big difference between M1, M2 and M3.
With M3MAX we get more CPU cores which are important for audio processing. This was not the case in the M1 and M2 generations, Pro and Max had the same CPU cores.

M3: 8 cores (4P + 4E)
M3: Pro 11 core CPU (5P + 6E) or 12 cores (6P + 6E)
M3: Max 14 core CPU (10P or 4E) or 16 cores (12P + 4E)

M1: 8-core CPU (4 performance cores and 4 efficiency cores)
M1: Pro 8- or 10-core CPU (6 or 8 performance cores and 2 efficiency cores)
M1: max 10-core CPU (8 performance cores and 2 efficiency cores)


M2: 8-core CPU (4 performance cores and 4 efficiency cores)
M2: Pro 10 or 12-core CPU (or 8 performance cores and 4 efficiency cores)
M2: Max 10-core CPU (8 performance cores and 4 efficiency cores)

Thanks, but you're not reporting if you had CPU overload on those loop points. So did you have these?

The difference, i e no overloads without loop points vs overloads with them, is the downside of testing with Live. It's a software error I believe, but very irritating.

Also, are the number of tracks where it always are no measurable errors, or does the exact possible amount of tracks vary?

i would interested to see if the m3 can get 20 tracks of the serum vst synth playing five-note chords, with an individual hall reverb on each track, at 256 buffer or smaller without audio breakup.

another test:
how many fabfilter pro-r reverbs can you run in series (meaning all on the same track) at a 256 buffer? this will test the single core performance, because running one reverb into another in a stacked single track fashion cannot be parallelized. if you can run ten pro-r reverbs on the same track (ten inserts, one after the other, one track), that would be impressive. i'm not saying you *would* run ten reverbs one into the other. i'm saying running several reverbs one into the other tests the system's ability to complete tasks within the dsp buffer.

If you (or anyone else) would not be "running several reverbs one into the other" then what use is this part of your test? Making artificial tests is both unnecessary and of little value, as I see it. May I suggest it would simply be more useful to mimic how most people would use the involved test candidates, perhaps using more returns for instance, if you're interested testing effects.

I feel acidrulez's choice with the demo song to be a pretty sound one. Of course one could also use something like DAWbench 2023 Benchmark Suite and increase track counts until crackling. This would necessitate migrating those to Live, if that was deemed important. Track count actually says a lot.

You should note that few (if any) DAWs would process effects on multiple cores. Most processing on a single track would take place on the same core. There are DAW differences, but far as I know only some instruments would spread processing (of additional notes) to other cores.

I've actually done your test on my M1 and found I could use a certain smaller number of a heavy hitting reverb plug-ins at the same time (Maybe 6-8 or so) before crackling ensued. Terrible, right? Well, I only use two such reverbs in my production so I never ever have overloads using those.

That does actually sound like a valid test. Just add some effects to it and you combine instruments and fx processing.

There's also another somewhat recent tester on Gearspace, that have done a lot of DAW load tests. But I haven't followed those posts for a while.

YouTuber James Zahn test added.

Hi, would you be able to do some testing in Pro Tools? I would be interested to know the performance difference between M1 Pro and M3 Max (16-core)

For context, my base M2 Mini Pro can play 18 tracks of Serum ( 5 note chord, factory patch ArcticWind ) with a hall on each track at a buffer of 256. CPU hit is around 60%. If I lower the buffer to 128, the CPU hit goes down to 55%. At any buffer size, the 19th instance of Serum with a hall reverb makes it glitch.

Interestingly, my CPU results were just marginally less without the hall reverb on each track and 18 was still my max of Serum tracks playing the chord patch.

Based on tests I've seen online, including the Zhan test referenced in this thread, the M3 will be less ( maybe significantly less...) than the M2 Pro Serum/Hall results.

I can do the test for you only on the M3Max if you want. I didnt order a M3pro. Are you still interested?

*definitely* interested in projects that are stress tested with serum, dense reverbs, and individual loaded with several cpu-heavy plugins in series.

100 tracks playing back relatively simple audio doesn't tell me much. a smaller project (16-20 tracks) loaded with serums, reverbs, group track processing, and master track processing tells me a lot.

Thanks for this - and ignore the negative comments, this is very useful to me.

Would it be accurate to say that despite the low scores in James Zahn's tests with the M3 Pro - the M3 Max having more performance cores means it beats out M1 Max and M2 Max?

Hey of course I'm happy if I can help.
There is an important difference between the M1,M2 chip and the M3 chip.
The M1 and the M2 chip had the same CPU cores in the pro and in the max.
The M3 chip has significantly more performance cores in the max.

To answer your question:

The M3max beats the M1max and the M2max.



Some useful Information: cores (performance+efficiency)

M3: 8 cores (4P + 4E)
M3: Pro 11 core CPU (5P + 6E) or 12 cores (6P + 6E)
M3: Max 14 core CPU (10P or 4E) or 16 cores (12P + 4E)

M1: 8-core CPU (4 performance cores and 4 efficiency cores)
M1: Pro 8- or 10-core CPU (6 or 8 performance cores and 2 efficiency cores)
M1: max 10-core CPU (8 performance cores and 2 efficiency cores)

M2: 8-core CPU (4 performance cores and 4 efficiency cores)
M2: Pro 10 or 12-core CPU (or 8 performance cores and 4 efficiency cores)
M2: Max 10-core CPU (8 performance cores and 4 efficiency cores)

I have a 12-core (8P+ 4E) M2 Pro.

Yes, I corrected it. Thanks for your help

Awesome and very helpful, I am likely going to order a 16" MBP M3 Max 16C 64GB and 2TB

A bit of a bummer that Pro Tools and Soundflow are not working fully on Sonoma yet - but I have been using both on my 13" MBP M1 with Sonoma for the last couple of weeks and workable for now.

In the new test I followed your suggestion and alternately loaded a track of serum and a track of Diva. Besides, there was still on every track
EQ Eight, ValhallaDelay,ValhallaVintageVerb,Compressor,multiband Dynamics, Glue Compressor and Archetype Nolly Guitaramp.

Serum and Diva were on the highest settings and played 6 notes at the same time.

Attached are the results.

44,1kHz
256 samples
Internal sound card

M1Pro 10 core: - 20 tracks
M3Max 14 core: - 36 tracks
M3Max 16 Core: - 38 tracks

That's impressive for the M3Max.

great test setup, thanks! very real-world (at least for me).

diva and serum are both cpu killers. so 20-38 tracks of serum/diva, reverbs and all sorts of other effects...256 sample buffer, internal soundcard (wow!)... that is definitely a great proving ground test.

Really interesting test thanks. I'm really surprised by the lack of difference between the 14 core and 16 core m3. 2 more performance cores give only 2 extra tracks. It doesn't seem worth it BUT if you want more than 36GB but don't want to pay for 96GB you're out of luck. The M3's are vastly superior to the M1 even though the M1 is still probably plenty good enough for most uses. The M1 Max and M2 Pro and M3 Pro are pretty similar in performance in their highest configurations as far as I can gather from youtube testing.

I'm still having a hard time deciding on how much RAM I really need. I'm thinking of doing some minimalist neo-classical / Icelandic style sample based orchestral library stuff but my focus has mostly been on vst and hardware synth based electronic music using very few samples. I'm tempted by 48 or 64gb mostly just for future proofing or if I did get more into the orchestral side of thing.

I do kind of want 4TB too even though I know I could get away with 2TB if I was selective and efficient. Increased Ram and Storage adds an extra £1000 though but when you're spending a minimum of £4000 why compromise - might as well spend £5000.