MacBook Pro M2 Max 12-Core 30 GPU or MBP M3 Pro?

[huzzam]

Originally Posted by Mt.Everest

TL;DR
Are you saying an M2 Max 12 core (w 8P cores) is the way to go vs an M3 Pro 12 Core (w 6P cores) ? ��

Yes. that's what I'm saying. You want as many P cores as you can get. Don't worry about *anything else* — GPU, E cores, or even the differences between P cores of different generations. (Each generation has a small improvement in single-core P-core performance; but more slightly-slower P-cores is better than fewer slightly-faster P-cores.) And don't worry about minor differences on the "compatible systems" list. That's an oversight on Avid's part.

Could you elaborate on this possibly? I may be missing the distinction btwn single and multi. Eg. If I have 1 track w 5 plugins, I expect I'll see 1 of the 8P cores in use. If I have a 2nd track, 5 different plugs, will this populate the next P Core or just add to P Core 1 till it's full?

To oversimplify: One track's entire processing chain will go on one core; another track's chain will go on another core. The OS will spread things evenly across the P-cores as much as possible. One core can have several tracks on it, until the total processing demand exceeds the abilities of that core.


Different tracks *can* go on different cores/threads, because their processing can be independent of each other. One track will always be on one core/thread, because each step of its processing is dependent on the previous step.


If you have 24 tracks, and each track requires 5% of a core for its processing, each of your 8 P-cores will be processing three tracks and will be about 15% in use. Again, this is oversimplifying but it's essentially how it works.

Then there's the AUX subs with plugs, Master fader, sends etc. Tbh, I don't know how single core vs multi works in PTs. The last time I thought of this was when maximizing my HD5 DSP slots available! Lol.

You're overthinking this. Aux tracks are just tracks. Master fader is just a track. Instrument tracks are just tracks. All of these will be processed like audio tracks.



Get more P-cores, and more than enough RAM for your usual sessions.

[Mt.Everest]

Originally Posted by huzzam

Yes. that's what I'm saying. You want as many P cores as you can get. Don't worry about *anything else* — GPU, E cores, or even the differences between P cores of different generations. (Each generation has a small improvement in single-core P-core performance; but more slightly-slower P-cores is better than fewer slightly-faster P-cores.) And don't worry about minor differences on the "compatible systems" list. That's an oversight on Avid's part.

ok got it! I believe this is the consensus from all previous comments as well. Thanks for confirming!

You're overthinking this. Aux tracks are just tracks. Master fader is just a track. Instrument tracks are just tracks. All of these will be processed like audio tracks.

Me, overthinking?? Never! (No need to look back at 20 years of my posts. Just trust me. lol)
But yes, i am well aware of this functionality of PT. I just thought something had changed with the way u first phrased it as relating to the Mx chips. Plus there's the occasional cold sweat I get when remembering the HD Core/Process DSP slots getting stuck in a reshuffling loop for 30 min while trying to print a mix after adding just onnne more plug. Haha. All good tho. I follow ya.

Get more P-cores, and more than enough RAM for your usual sessions.

So I believe this thread is complete and I really appreciate all the help from everyone.

One last thing I wanted to note was that the reason the MBP choice was so specific was that I had already purchased a Refurbished MacBook M2 Max 12-Core, 32GB RAM, 1TB SSD. It was the only viable option at the time compared to the M3 Pro. (Both details listed above).

So, I didn't have the option to upgrade the RAM (unified memory, whatever) as the refurb came with 32GB. I haven't opened the box yet and while I'm 99% sticking with it, do you think I overlooked the importance of bumping the RAM up to 64GB which I could have done w a new machine?

I'm coming from a 16GB 2013 Intel MBP, and figured 32GB would be good enough. Sessions are generally <64 tracks, <150 plugs, and any VIs are used in the early stages but disabled during mixing.
I should get by ok with 32GB, no?

Thanks all!
Cheers

[LDS]

Has anyone run a similar test with 2023.12? I wonder if Pro Tools may receive an update at some point in time that allows it to access the E-cores in the same way Reaper or Nuendo does.

[uptheoctave]

I've got an M3 Max with 16 cores (12 performance and 4 efficiency), 64GB RAM, 2TB and it is ridiculously fast.

See attached benchmarks.
It is faster than the M1 Ultra and by a considerable amount in terms of CPU.
It is approaching double the speed of the 2021 M1 Max that it replaced.

[LDS]

Originally Posted by uptheoctave

I've got an M3 Max with 16 cores (12 performance and 4 efficiency), 64GB RAM, 2TB and it is ridiculously fast.

See attached benchmarks.
It is faster than the M1 Ultra and by a considerable amount in terms of CPU.
It is approaching double the speed of the 2021 M1 Max that it replaced.

The murkiness comes from Apple's decision to leave P & E core utilization up to each developer. Reaper & Nuendo use both P & E cores, so their performance closely resembles the geekbench scores. Pro Tools & Logic Pro on the other hand only use P cores. The resulting real world performance seems to be 'inverted' compared to the Geekbench results because the E cores are taken out of the picture - M3 Pro has lower P core performance than the M2 Max, and the M2 Max has lower P core performance than the M1 Ultra.

[uptheoctave]

Originally Posted by LDS

The murkiness comes from Apple's decision to leave P & E core utilization up to each developer. Reaper & Nuendo use both P & E cores, so their performance closely resembles the geekbench scores. Pro Tools & Logic Pro on the other hand only use P cores. The resulting real world performance seems to be 'inverted' compared to the Geekbench results because the E cores are taken out of the picture - M3 Pro has lower P core performance than the M2 Max, and the M2 Max has lower P core performance than the M1 Ultra.

Hi LDS, oh yeah, I agree.

Even weirder that Apple chose not to fully utilise their own hardware by software they actually make themselves.

My main issue is how Apple report their internal benchmarking to be incredible performance gains generation to generation, when the reality is slightly more modest. It seems weird to me, like a dating profile where someone uses a picture 10 years out of date. I mean, your eventually all know what the person looks like/
how the machine performs. Better to not exaggerate.

I should be able to compare the performance in PT and Logic between the Studio (M1 Ultra) and MBP (M3 Max).

I guess Avid will address this issue over time, or at least I hope so.

I tend to think any of the higher powered M1/2/3 machines are fine for almost anything audio related. I didn't upgrade from an M1 Max to an M3 Max because I needed the power. We just turn the machines over ever 2-3 years as part of a business refresh.

The M3 Ultra should be an absolute weapon when it comes.

[knowtice]

If it helps your decision at all I'm using an M1 Macbook Pro, 8gb ram, 512mb drive, and have PT2023.12 current session is 73 tracks, SSL channel on every track, lindell 50's, oxford limiter, delays, etc and I'm using about 50% cpu and the machine is not working hard at all. In fact the session is open as I'm typing this. Bottom line is I think it's getting to the point where everything available is more than capable.

[[email protected]]

Originally Posted by knowtice

If it helps your decision at all I'm using an M1 Macbook Pro, 8gb ram, 512mb drive, and have PT2023.12 current session is 73 tracks, SSL channel on every track, lindell 50's, oxford limiter, delays, etc and I'm using about 50% cpu and the machine is not working hard at all. In fact the session is open as I'm typing this. Bottom line is I think it's getting to the point where everything available is more than capable.

My setup is similar.. 14" M1 MBP. I'm really not mixing projects on my MBP, I'm mainly tracking when I'm remote like that. I've got Carbon/Carbon pre for 16 Channels and can go another 16 at 44.1 if I need it but usually roll 88.2 when I'm 24 channels via ADAT. I'm recording big ensemble stuff, a lot of stereo mics etc..

I've never had a hint of hiccup on this machine. Recording straight to the internal drive.

[LDS]

Originally Posted by uptheoctave

Hi LDS, oh yeah, I agree.

Even weirder that Apple chose not to fully utilise their own hardware by software they actually make themselves.

My main issue is how Apple report their internal benchmarking to be incredible performance gains generation to generation, when the reality is slightly more modest. It seems weird to me, like a dating profile where someone uses a picture 10 years out of date. I mean, your eventually all know what the person looks like/
how the machine performs. Better to not exaggerate.

I should be able to compare the performance in PT and Logic between the Studio (M1 Ultra) and MBP (M3 Max).

I guess Avid will address this issue over time, or at least I hope so.

I tend to think any of the higher powered M1/2/3 machines are fine for almost anything audio related. I didn't upgrade from an M1 Max to an M3 Max because I needed the power. We just turn the machines over ever 2-3 years as part of a business refresh.

The M3 Ultra should be an absolute weapon when it comes.

All of the benchmarking stuff is starting to remind me a bit of the VW diesel emissions saga. As long as it initially tests well, it’ll sell. Apple, Google and Samsung all embargo benchmark results on early reviews. It wouldn’t surprise me if they were spending that time fine tuning performance to better the benchmark results before they are made public.

That’s not to say they aren’t great machines. They are. It just seems like the whole performance gain/benchmark angle has woven itself into the marketing as a massive priority.

[JFreak]

Nah, it's just benchmarking multicore with offline processing. Really useful metric for PT