Ableton Forum

Hello Ableton community.

I'm on the cusp of upgrading to a Mac Studio, from a 2013 Mac Pro. I'm currently deciding between the expensive M1 Max, or the absurdly expensive M1 Ultra.

To help with my decision, I'm hoping some of you here could provide some technical insight into the nature of performance bottlenecks in very large Ableton projects.

My projects at the moment involve a lot of experimental sound design, I often end up with 200/300 tracks in a single project.

Not playing simultaneously of course, I tend to freeze / resample and work with audio as much as possible. Still, I like to keep the patches to hand.

I've noticed that even when 99% of these tracks are frozen, performance is still heavily impacted. The interface becomes sluggish, adding / removing devices causes hangs, editing automation becomes laggy and unresponsive.

At a certain point, I'm forced to spend time cleaning up the project, saving out patches to my library then removing them from the set. Then things become smooth again, for a while at least.

So even when tracks are frozen, tucked away in the project and not being played, some resources are still being used, affecting performance.

Getting back to my question:

As I understand it, the only thing M1 Ultra offers over M1 Max is a theoretical doubling in multithreaded performance.

Single core speeds are identical between the two systems, as are other factors such as memory and storage speed.

This has me wondering how much improvement I'm likely to see with the increased multicore performance, particularly with regard to fluidity and responsiveness.

I'm concerned the bottleneck might be taking place elsewhere. Single core speed? Memory bandwidth? Storage Speed?

If any of you could offer insight on this topic, I would really appreciate it!

Thanks,

Joe

Hey Joezay.

I'm looking for the same answers.
I watch a lot of test and comparison but nothing concerning audio (except logic) and nothing of course of the behaving of Live in an Max or Ultra Studio.
& I know that Live use CPU and other stuff from the computer differently then Logic f.e.
& I know also that the Ultra version - depending of the software - is not "doubling" the performances.
So not sure that Ultra (comparing to the price) worth it.
Do you have more infos about it now ?

Cheers

Difficult to answer without actually testing your Live sets on the computers you want to buy...

I come from a mid-2012 macbook pro with an i7 and 16Gb of RAM. I bought recently a 16 inches M1 pro with 32gb of RAM.
I don't have any more the problems I had before, but of course it is kind of obvious.

What I noticed after more than one month of use :
- my brand new M1 pro (macbook pro 2021 16 inches) is between 2 to 3 more efficient than my old i7 mid-2012 macbook pro when run in Rosetta mode. It means I can run 2 to 3 more plugins/tracks than before (depending on the plugins - see below). It's not spectacular, but of course it's still much better.
- when using it in Native, it adds around 25% more of CPU headroom compared to Rosetta.
- I found that what truly affects performances are the plugins I use in a project. Some are very well coded for Apple Silicon (like the last Arturia updates) and it seems you can't make the CPU peak even if you use hundred of such plugins (I tested it). Some are not that great and I don't find as much improvement with them compared to my old machine. Some are still not even Apple Silicon Native (like Native Instruments, Sonimus, Reason, SSL, Boz digital labs...) and can make the session unstable.
- I run sometimes very large Live sessions, and on my old machine the bottleneck was not only the CPU but also the RAM. The 32 gb of very quick RAM helps a lot...

Maybe you could share with us a session you don't care sharing the intimacy of it, and we could try to run it on our machine, so we can tell you if it has stil issues or no.

Theoretically I don't think the RAM or the SSD speed will affect the performance : they are insanely quick and efficient in my new macbook. It's night and day compared to my old laptop's RAM and 3d party Samsung SSD.

But you need to check how much RAM you use in your sets, and choose a machine with more RAM than that : don't believe the tales about running crazy sessions with 8Gb of RAM. It's working great because the SSD is very quick so swapping is not so bad as before, but it's much better with the correct size of RAM. And if you freeze a lot of tracks or use a lot of samples, RAM is important...

So I guess the only bootleneck for you will be the multi-threading efficiency and the plugins you use...

Thank you for the responses.

So after posting this I actually bought an Ultra with 128GB of memory.

The system performed well as expected, but I had to return it due to major issues with the display output. The Mac Studio caused severe flickering and persistent image artifacts on several different displays. There seems to be problem with monitor compatiblity on the M1 platform which Apple is refusing to acknowledge:

https://forums.macrumors.com/threads/m1 ... 670/page-3

I had such a bad experience with Apple here, I decided to move my entire production workflow to a custom built PC.

My new build has a 12900K, 64GB DDR5, Gen4 PCI SSDs. I'm running Windows 11.

On paper performance is roughly equivalent to the Studio Ultra, and built for less than half the cost.

So far the PC has been rock solid, extremely stable. Two weeks in and zero crashes so far (my old Mac Pro would crash up to three times per session).

A bit off topic, but here's my outline of some pros and cons as compared to the Mac Studio:

Pros:
Fully user serviceable and upgradable over time (huge one for me).
The display output actually works properly.
Better price to performance ratio.

Cons:
Lots of tweaking required in Windows to optimize audio performance (drivers, DPC latency, etc).
No more Time Machine. Built in backup systems on Windows are unfit for purpose, I went with Crashplan ($10 monthly cost).
Windows Explorer is inferior to Finder (only able to view contents of 1 folder at a time).
Keyboard shortcuts are different (years of ingrained muscle memory to retrain).
Increased noise and power consumption (energy costs may be significantly higher over the life of the machine).

turns out that for either mac or pc, the answers lie with JUCE and ABLETON... and which one to blame, which one is not moving an inch to solve these excruciating issues when it comes to large sets.

https://www.reddit.com/r/ableton/commen ... i_remains/