Logic Pro - m4 max or m3 ultra?

[Royal Gala]

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

I tend to work with hundreds of tracks (both software instruments and audio) with cpu intensive plugins (at least a good 10 or more plugins per track). Because of bad habits, i don’t really tend to work with bus tracks (so I might have multiple individual travjs with the same individual plugins doing the same thing, where they could have been bussed to a single track). Would you still say higher single core speed is preferable over multi core in my case?

Thank you to everyone for the replies - this is all super helpful.

 

[streetfunk]

. Would you still say higher single core speed is preferable over multi core in my case?

I´m not the person to answer on that.
My workflow is entirely different.


I personally would allways grab the latest CPU tech with highest single core speed.
But you are the pro on your work.
My work looks really completly different.

edit: M3 to M4 seems to be a 25% increase in SC speed. That´s allready quite a nice jump.
M2 to M4 is from what i saw with my work, 40% if not even more. Probably even close to a 50% increase vs. my setups. So, in my opinion not anything smaller than Geekbench 6 was suggesting it to be.

 

[zedsdead]

for 98% of users and professionals the M4 Max will be better due to the higher single core clock speed. There is still plenty of RAM available on the Max chipset.

I know it's video production, but Severence was edited on Intel Mac mini's and with newer Macs through Jump.

For example, my 4k video workflows....while I myself have a M2 Mac Studio - I am sure it is overkill....at this point I only need a decent amount of RAM...any of these chipsets are more than powerful enough. At the time of purchase, the Mac mini was still in the old design and RAM limits pushed the price up too high, so the studio was simply such a better value. I also have an M1 Max MacBook Pro.

Next time I need to upgrade, I am seriously going to consider "lower end" machines.

This idea applies to the high end also.

If you really think you will utilize the extra RAM or bandwidth, then get the M3 ultra. But few people actually will need it.

 

[krell100]

It's a pity Logic is kind of hamstrung by not utilising efficiency cores at all, as other DAWs (Reaper, Cubase etc) that are able to use 'all' cores (p + e) see much higher performance overall as far as number of tracks. Of course this won't help on high load single tracks but for those with huge track counts it's something to consider.

 

[Mr_Time]

It seems the general assumptions of M3 Ultra vs M4 Max are being made primarily on the Geekbench 6 results. There is no arguing that while the M4 is indeed a faster chip than the M3, digging a little deeper into the single- and multi-core results start to paint a slightly different picture when it comes to applying those benchmarks to Logic.

1. For single-core, the often mentioned speed increase of ~23% is due in part to the three sub-tests in the single-core Geekbench 6 tests that utilize SME. As the M3 doesn’t have SME, those results skew the overall average for the M4. One test in particular, “object detection”, is over 200% better on the M4. When you remove the 3 tests from the average, the results are closer to an aggregate ~12% over the M3.

SME is used for on-device object detection, photo library classification, and background, especially in mobile:

• Photo capture: face recognition, scene recognition, object tracking
• Security: facial recognition
• Videoconferencing: background blur, object detection
• Photo library classification
• Apple's "remove background" feature

Why is this important when deciding on a machine for Logic? Because none of those use cases are relevant to Logic nor other audio applications. Great for finding that pic of your dog in photo library, but not relevant for Logic.

When looking at some of the Photoshop CPU only benchmark tests, such as from Artisright, the M4 Max is faster but only by a smaller margin in line with the generational step from M3 to M4, not a 25% increases as the Geekbench 6 results show, as SME is not at play there.

2. Conversely, for the multi-core Geekbench 6 benchmark, the real number is actually much higher than the 8% difference for the M3 Ultra over the M4 Max. The Geekbench 6 multi-core test has already been determined to be inaccurate, as Geekbench 6 is not built to test large core counts properly…

Don’t trust those early Geekbench M3 Ultra benchmarks just yet

Most real world multi-core benchmarks, in video for example, are showing ~40% gains. The Music-Prod Logic benchmark track score of 552 tracks is also an indication that the gain is more than +8% over the M4 Max.

Again, while the M4 is indeed a faster chip, the real world benefit when it comes to Logic should not be interpreted solely from the synthetic Geekbench 6 results. The often cited +25% single-core advantage for the M4 Max and and only +8% multi-core advantage for the M3 Ultra shouldn’t be accepted blindly.

For the purposes of deciding on either the M3 Ultra or M4 Max machine for Logic, those number are probably closer to +12% single-core M4 Max and +40% multi-core M3 Ultra when you take SME out of the equation and look at Geekbench’s inability to properly test large core counts.

Somehow, this makes the decision even harder to make 😂

 

[cl516]

Good thread here. I went with the M3 Ultra base model on Friday, should be great for my workflow. Will need a few days to get it up and running. Having a few bumps in the road with Migration Assistant from M2 Max Studio, but should be fine.

 

[cl516]

Mac OS 15.3.2. M3 Ultra, 28 core, 96GB ram, 1TB. Logic 11.1.2. Session at 96khz sample rate. Ok friends I have some info:

Using the Diva patch described earlier, “I Can’t Believe It’s Not Analog”. Accuracy set to Divine. Multicore: ON. Poly 16 voices. Stack 2. Logic I/O 64 samples. Processing Threads: Automatic. Process Buffer: Large. With these exact settings, I can play the full 16 voices live, with long tails, zero crackling. Not even a hint. Logic’s CPU meter looks like it’s having a heart attack, on one single core. Just completely 100% on a final thread in the meter.

I got crackling if I: Turn OFF Multicore in Diva. Lower Process Buffer Range. And surprisingly, 256 I/o buffer size wasn’t as agreeable as 64.

And if the MIDI part is already recorded, and I don’t highlight the Audio Instrument channel, then Logic immediately knows to allocate the workload to the other processing threads. In this case with the specific Diva example above, Logic pushed it over to the first 3 processing threads, each hitting about 40%.

There’s so many other variables though, and this is only 1 single soft synth. Different instruments / effects will consume CPU differently. I hope somebody finds this info helpful. Good luck all!

 

[cl516]

P.S. Tripling the number of parts from Diva (identical part), that only increased the Logic thread usage from 3 to 5 bars, with 3 of the first bars increasing up to just over 50% usage.

Got to the point where I had 10 softsynths playing back live, each with pretty heavy usage. Logic knows to keep offloading the work to the next thread. 6 threads remain empty. And the ones that ARE doing work, they seem to be capped at 75% usage. 🤷🏻