M4 Pro vs. M5 MacBook Pro Buyer's Guide

[StuBeck]

This is exactly the challenge that I'm having - real real world benchmarks.

My partner has a 2019 MBP, Intel, 2.6GHz 6-core i7. She doesn't game, render 8K video, etc, but, does work with quite large inDesign and similar documents for her work.

As a bit of a test, she opened a 800 page document, filled with complicated layouts, embedded images, etc. She then exported it to PDF on the current MBP. It took about 7 minutes (slightly over). I should point out that some of her documents are multiple thousands of pages long, so this was a relatively small example.

The same file on her home MacStudio, M1 (not pro/ultra) with the same task took just under 4 minutes. On the MacStudio, I started up Activity Monitor and watched each core and the GPU during the export.

Whilst not super scientific, it was interesting to see that on the MacStudio, it only used about 4 of the performance cores, and didn't max them out. The GPU(s) was probably about 20-25% busy, from the graphs.

This is the sort of real-world benchmark that has meaning for someone like her. The time she can't really use her computer because it's busy doing this task, and there is a time criticality to it. She often has to have these massive documents processed by a "hard stop" time that can not be changed (down to the minute), and, sadly, due to the role of other people, it's often not possible to simply "start sooner" to have time to export the document before a deadline.

She's vacillating between the new M5 and the M4 Pro. Why not wait for the M5 Pro? Well, this is a corporate purchase, and funding cycles are what they are. She gets one laptop refresh every four years, and if she doesn't act in the next few weeks, it's not clear when she may be able to get one.

She doesn't use the screen on her laptop as she has two monitors at work connected, so the 14" vs 16" isn't really a factor.

It's not clear whether the "AI/Neural" hardware has any real impact, but, that may depend upon what Adobe does. And no, she can't change to some other software. She likes inDesign, and there is a significant workflow involving dozens of people that is involved (large organisation).

My sense is that the M5 would be more than enough for the next four years for her, as it will absolutely be better than the M1 MacStudio, and significantly better from her Core i7 MBP.

Curious if others have similar thoughts or not...

Is it a Mac Studio or a Mac mini? The studio only came in Ultra/Max configurations. If she went to an M5 laptop I would assume she would see a slight decrease in speed.

I personally went from an M1 Pro to an M4 Pro and saw an improvement, but it wasn't like I was suddenly doing things in 3 minutes which took 10 before.

 

[UrfinJuice]

Nothing you’re doing in music production should bring that processor anywhere close to “its knees”. It sounds like you’re most likely hitting the effects of impressively bad (unoptimized) software. Something jre-based, perhaps?

edit: fixed typo

Yeah you´re right, nothing should. Unfortunately, something does though. With a handful of synths playing the same chord simultaneously, or even just one synth, you can overload the M4M CPU. And that's before even using any effects. Undoubtedly, the M CPUs are a huge step forward, but there's still a lot of work to be done regarding single-core performance before users like me can truly benefit.

 

[appeals.key_0k]

Yeah you´re right, nothing should. Unfortunately, something does though. With a handful of synths playing the same chord simultaneously, or even just one synth, you can overload the M4M CPU. And that's before even using any effects. Undoubtedly, the M CPUs are a huge step forward, but there's still a lot of work to be done regarding single-core performance before users like me can truly benefit.

That’s just it though. What you’re describing isn’t bringing the M4 to its knees. This sounds like a developer refusing to parallelize their application. The power is there; the responsibility is on the application developer to use it. We reached diminishing returns on single-threaded performance quite some time ago.

Conceptually, it would be similar to me using 32bit addresses in my application, inherently limiting my memory usage to ~4gb, then claiming that the 16gb (or 32gb, 64gb) of memory in the system isn’t good/fast enough. Sure, we could increase the memory speed and bus bandwidth, but at a point, it’s more economical to modernize the application to use more of the available resources.

edit: just guessing here, of course. Would need to see the per-core resource usage to know for sure.

 

[artifex]

Yeah you´re right, nothing should. Unfortunately, something does though. With a handful of synths playing the same chord simultaneously, or even just one synth, you can overload the M4M CPU. And that's before even using any effects. Undoubtedly, the M CPUs are a huge step forward, but there's still a lot of work to be done regarding single-core performance before users like me can truly benefit.

What's your DAW, and have you tried a trial version of another one? Also, are these software synths you're using when you're crashing? Is any part of your toolchain pre-Silicon, including effects that you aren't using, but still get loaded? Make sure everything is ARM-native VST or AU, not some old Intel-compatible VSTs from a decade ago that have to use Rosetta 2, at least just for testing.