Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I’m also debating between the M3 Ultra and M4 Max Mac Studio for Logic Pro.

Right now I don’t need more than 96GB RAM, so looking at either the M4 Max 16/40 128GB 1TB or the M3 Ultra Base. Roughly the same price. (In fact the M3 Ultra Base can be found for less through 3rd party dealers). So when price is taken out of the equation, it becomes even a harder decision.

I know that both will serve my purpose, but does the better single performance, and newer chip family of the M4 Max outweigh the better multi-core (more cores), additional GPUs, and memory bandwith for Logic Pro in the M3 Ultra? Plus you get an additional two TB5 ports.
 
Me three. Same question as you guys. "Conventional Wisdom" online seems to lean towards the M4Max, but for my workflow maybe more performance cores in the Ultra is better. Currently undecided. 🤷🏻
 
Remember how DAWs allocate the CPU ressources dependend on how you deal with AUX- and Bus- Tracks and master inserts etc. I´d say, the higher the single core speed the better. Anyway.
But overall, that´s finally all usecase dependend.

Ask yourself how much RAM you need to have.
The ultra is about the RAM, and for some things about the graphics power.
Is any of this relevant to your audio work ?*

* there´s your answer ;)
 
The M4U has 20 p cores versus 12 p cores in the M4M 16/40 so I guess the ultra would suit high track counts, but the M4s higher clock would probably suit a mixing/mastering session more with a long chain of plugins on one core. Tricky…
 
The M4 Max is historically the most powerful Mac except for the M3 Ultra, both in single core and multicore (slightly edging out the M2 Ultra). As mentioned, track count is only part of the problem. Inserts, sends, groups and master bus can significantly change the scenario. As an example, I have a test project that consists of dozens of tracks of U-he Diva & Repro synths. With the quality set to the highest level and more CPU-expensive patches loaded, My M4 Max Studio kills my old M2Ultra because the higher core performance of the M4 allows better utilization of the 12 p-cores than the M2 can of the 16 p-cores. On the other hand, if you have a project with a couple hundred lightweight tracks (from a CPU standpoint), the M2 Ultra was just as performant as the M4 Max.

I have been doing music production on computers a very long time and normally work with large sample-based orchestral libraries and synth's such as Omnisphere, Diva, Dune, etc. Track counts on finished projects are 100+ tracks (sometimes way more). 128GB is the minimum RAM I would consider. In spite of that, I chose the M4 rather than M3 because core spiking is usually the limiting factor, not running out of cores.

I think anyone who is not primarily an orchestral composer working with 300-500 track templates of sample based instruments should not consider the M3 Ultra. And if you are a M3 Ultra candidate for music, you will need more than 96GB of RAM because you are probably sample based.

DAW's don't parallelize like a Final Cut Pro video workflow does. The M3 Ultra may be 50% "faster" on multicore benchmarks than the M4 Max, but that often won't be realized in audio work. If you currently are working on an M2 Ultra machine (or maybe a M3 Max) and are frequently getting overloads in Logic, Cubase, etc, then a M3 Ultra may be the answer, but maybe not. Frankly, it is difficult for me to imagine a specific DAW workflow/project that will take advantage of an Ultra with the base 96GB of memory as most of the really high track count use cases (hundreds, not dozens) are using huge sample libraries that also gobble up RAM.
 
The M4 Max is historically the most powerful Mac except for the M3 Ultra, both in single core and multicore (slightly edging out the M2 Ultra). As mentioned, track count is only part of the problem. Inserts, sends, groups and master bus can significantly change the scenario. As an example, I have a test project that consists of dozens of tracks of U-he Diva & Repro synths. With the quality set to the highest level and more CPU-expensive patches loaded, My M4 Max Studio kills my old M2Ultra because the higher core performance of the M4 allows better utilization of the 12 p-cores than the M2 can of the 16 p-cores. On the other hand, if you have a project with a couple hundred lightweight tracks (from a CPU standpoint), the M2 Ultra was just as performant as the M4 Max.

I have been doing music production on computers a very long time and normally work with large sample-based orchestral libraries and synth's such as Omnisphere, Diva, Dune, etc. Track counts on finished projects are 100+ tracks (sometimes way more). 128GB is the minimum RAM I would consider. In spite of that, I chose the M4 rather than M3 because core spiking is usually the limiting factor, not running out of cores.

I think anyone who is not primarily an orchestral composer working with 300-500 track templates of sample based instruments should not consider the M3 Ultra. And if you are a M3 Ultra candidate for music, you will need more than 96GB of RAM because you are probably sample based.

DAW's don't parallelize like a Final Cut Pro video workflow does. The M3 Ultra may be 50% "faster" on multicore benchmarks than the M4 Max, but that often won't be realized in audio work. If you currently are working on an M2 Ultra machine (or maybe a M3 Max) and are frequently getting overloads in Logic, Cubase, etc, then a M3 Ultra may be the answer, but maybe not. Frankly, it is difficult for me to imagine a specific DAW workflow/project that will take advantage of an Ultra with the base 96GB of memory as most of the really high track count use cases (hundreds, not dozens) are using huge sample libraries that also gobble up RAM.
You make some very good points based on real world use cases to go for the M4 Max.

I happened to come across this recent article which also expresses a lot of the same rationale, however when he gets to the recommendation he seems to choose his words carefully.


He concludes:
“My advice when it comes to audio and music production is to look for CPUs which near the top of the single thread/single core charts — as that’s the most likely choke point for modern processors — and then look for the best overall (multicore) performance from the CPUs in that group.”

He specifically says “CPUs which near the top”, not “the fastest CPU” in single core performance. So based on this, both the M4 Max and M3 Ultra are “near” the top in single core charts, with them being #1 and #2 respectively for the Mac range. But then taking the second criteria into consideration, “best overall multicore performance”, that would lead you to the M3 Ultra based on his logic.

I realize it’s use case dependent, as you say, but his general recommendation seems to favor a reasonably fast single core but the fastest multicore.
 
  • Like
Reactions: cl516
You make some very good points based on real world use cases to go for the M4 Max.

I happened to come across this recent article which also expresses a lot of the same rationale, however when he gets to the recommendation he seems to choose his words carefully.


He concludes:
“My advice when it comes to audio and music production is to look for CPUs which near the top of the single thread/single core charts — as that’s the most likely choke point for modern processors — and then look for the best overall (multicore) performance from the CPUs in that group.”

He specifically says “CPUs which near the top”, not “the fastest CPU” in single core performance. So based on this, both the M4 Max and M3 Ultra are “near” the top in single core charts, with them being #1 and #2 respectively for the Mac range. But then taking the second criteria into consideration, “best overall multicore performance”, that would lead you to the M3 Ultra based on his logic.

I realize it’s use case dependent, as you say, but his general recommendation seems to favor a reasonably fast single core but the fastest multicore.
The problem is the M4 core is significantly faster than the M3 core. M3 was a more incremental improvement over M2
 
I have a M2U 128 GB 2TB installed here. Was lusting over the M3U and even the M4M. In the end I decided to wait another release cycle of either a M5M or a M4U. The M3U (10% SC, 30% MC) and the M4M (20 SC, 10% MC) isnt that much of a gain on a even not 2 year old machine. I can get 4K back for the M2U but must invest another 2.5K to (6.5K-4K)upgrade. I think it is not worth the hassle and sit this one out.
Screenshot 2025-03-26 at 13.52.30.png
 
I have a M2U 128 GB 2TB installed here. Was lusting over the M3U and even the M4M. In the end I decided to wait another release cycle of either a M5M or a M4U. The M3U (10% SC, 30% MC) and the M4M (20 SC, 10% MC) isnt that much of a gain on a even not 2 year old machine. I can get 4K back for the M2U but must invest another 2.5K to (6.5K-4K)upgrade. I think it is not worth the hassle and sit this one out.View attachment 2495956
I think you are underestimating the single score gain of the M4 over the M2 - it's more like 40% based on most benchmarks I've seen.
 
  • Like
Reactions: streetfunk
yeah you are right 4000/2800=42%
Looks a bit better yeah, but still I will wait it out for a M4U or M5

I will lose the better heatsink going to a Max instead of Ultra, also I need to boost the MEM to 128GB (I have that now as well) and then it is quite an investment for a second choice macstudio (M instead of U) and my Studio is bought in June23, so still more than 1 year warranty.

I really want to go from 48khz to 96 Khz in my templates, that needs a lot of juice.

Also I always want some headroom on my single core, due to spiking. So it is M4 or better and no M3.
Really disappointed Apple didn't release the M4U, that would be an instabuy.

Believe me I had the different configs in my basket and came to the conclusion I wil wait 1 cycle.
In my mind I will OCD on the 2 lost TB ports and weaker heatsink when going for the M4M.
 
yeah you are right 4000/2800=42%
Looks a bit better yeah, but still I will wait it out for a M4U or M5

I will lose the better heatsink going to a Max instead of Ultra, also I need to boost the MEM to 128GB (I have that now as well) and then it is quite an investment for a second choice macstudio (M instead of U) and my Studio is bought in June23, so still more than 1 year warranty.

I really want to go from 48khz to 96 Khz in my templates, that needs a lot of juice.

Also I always want some headroom on my single core, due to spiking. So it is M4 or better and no M3.
Really disappointed Apple didn't release the M4U, that would be an instabuy.

Believe me I had the different configs in my basket and came to the conclusion I wil wait 1 cycle.
In my mind I will OCD on the 2 lost TB ports and weaker heatsink when going for the M4M.
Still not certain what type of production you are doing. There isn't much point to running at 96k for sample based orchestral stuff since most sample libraries are 48k. Yes the M4 Max runs hotter when really pushed, but the fans aren't ramped up much and are still silent in my experience. I sorta ignore CPU temps - if Apple thinks it's safe to run there, who am I to argue? And AppleCare+ is only $59/year for as long as you own the machine, so if it does eventually cook itself, I'm not out $$$$$.

I do agree a M4 Ultra would have been a slam dunk for me, regardless of cost. The M4 Max vs M3 Ultra wasn't so much a cost thing as a single core speed thing. If they deliver a M5 Ultra in a couple years, I'll jump on that and turnover the M4M.
 
Apple M4 Is The Fastest, But Not Really
It's a video that's mostly playing to the Apple haters (see the comments on the video), but the point that synthetic benchmarks cannot tell you how your use will be is still a valid point.
 
It would be interesting to see how an M4 based mac goes with one track with a lot of tough plugins on, say a synth like Diva or Pigments with a very intense preset going in to a high end delay and then a high end reverb, then a compressor etc etc
 
  • Like
Reactions: cl516
I have been doing music production on computers a very long time and normally work with large sample-based orchestral libraries and synth's such as Omnisphere, Diva, Dune, etc. Track counts on finished projects are 100+ tracks (sometimes way more). 128GB is the minimum RAM I would consider. In spite of that, I chose the M4 rather than M3 because core spiking is usually the limiting factor, not running out of cores.

I think anyone who is not primarily an orchestral composer working with 300-500 track templates of sample based instruments should not consider the M3 Ultra. And if you are a M3 Ultra candidate for music, you will need more than 96GB of RAM because you are probably sample based.
I had the same thoughts, and just received my 128GB M4 Max. Fingers crossed it works out nicely. Wish it would have been possible to get 192GB RAM though, that is my main concern; running out of memory.
 
  • Like
Reactions: cl516
I had the same thoughts, and just received my 128GB M4 Max. Fingers crossed it works out nicely. Wish it would have been possible to get 192GB RAM though, that is my main concern; running out of memory.
Well if you get the chance, I'm sure we'd all love to see some CPU meter usage on running stuff like Diva / Pigments / Serum with the M4 Max! Would love to hear more about this. 🙏
 
Well if you get the chance, I'm sure we'd all love to see some CPU meter usage on running stuff like Diva / Pigments / Serum with the M4 Max! Would love to hear more about this. 🙏
Diva, accuracy=Divine, MC=off, patch"I Can't Believe It's Not Analog"

128 buffer (UFX III) playing 4 & 5 voice chords with pedaling in between (so voice count is even higher)

Logic 11.1.1, Diva track record armed

M4 Max - single thread at about 75% worst case per the Logic performance meter. Activity Monitor displays significantly less physical core loading

M2 Ultra - immediately overloads on the first chord
 
Diva, accuracy=Divine, MC=off, patch"I Can't Believe It's Not Analog"

128 buffer (UFX III) playing 4 & 5 voice chords with pedaling in between (so voice count is even higher)

Logic 11.1.1, Diva track record armed

M4 Max - single thread at about 75% worst case per the Logic performance meter. Activity Monitor displays significantly less physical core loading

M2 Ultra - immediately overloads on the first chord

For me on my current M2 Max Mac Studio, OS 15.3.2, Apogee Symphony I/OMk2
Logic 11.1.2, 64 i/o buffer, 96khz session

Diva, accuracy=Divine, MC=off, patch "I Can't Believe It's Not Analog"
I can get a 4 note chord on that patch. (Which is already set to Stack2)
Anything more, and it instantly crackles.
 
  • Love
Reactions: krell100
For me on my current M2 Max Mac Studio, OS 15.3.2, Apogee Symphony I/OMk2
Logic 11.1.2, 64 i/o buffer, 96khz session

Diva, accuracy=Divine, MC=off, patch "I Can't Believe It's Not Analog"
I can get a 4 note chord on that patch. (Which is already set to Stack2)
Anything more, and it instantly crackles.
Tried that on my M4 Max - while the Logic performance meter shows the live thread at 100%, I am able to play 10 voices at once and play 4 and 5 voice chords using the sustain pedal in between without overload or crackles.
 
  • Like
Reactions: cl516 and krell100
Tried that on my M4 Max - while the Logic performance meter shows the live thread at 100%, I am able to play 10 voices at once and play 4 and 5 voice chords using the sustain pedal in between without overload or crackles.
Would be great to see this test on the M3 Ultra, just to compare.
 
  • Like
Reactions: cl516
I set up the test with Diva on my 2013 MacPro 3.5Ghz 6 core (Intel Xeon E5) just for amusement. OS 12.7.6.
Logic 10.7.9, 128 i/o buffer, 44.1khz session. This is going in to my Arturia Audiofuse 8Pre.

Diva, accuracy=Divine, MC=off, patch "I Can't Believe It's Not Analog"
4 note chord it's around 90-95% of one core.

1743227447671.png
OS.png
 
  • Like
Reactions: cl516
I set up the test with Diva on my 2013 MacPro 3.5Ghz 6 core (Intel Xeon E5) just for amusement. OS 12.7.6.
Logic 10.7.9, 128 i/o buffer, 44.1khz session. This is going in to my Arturia Audiofuse 8Pre.

Diva, accuracy=Divine, MC=off, patch "I Can't Believe It's Not Analog"
4 note chord it's around 90-95% of one core.

View attachment 2496760View attachment 2496761
Interesting - I had a 6 core MP 5.1 and I was constantly fighting single core overloads. Maybe the modern Diva is really well optimized on X64 these days. My old MP is long gone but that is not at all what I would have expected.

If I just play 4 notes, my M4 Logic meter shows about 45%
 
Last edited:
  • Like
Reactions: krell100
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.