M3 Pro chip a bad choice for Logic Pro?

[Richard Herczeg]

I would love it if Apple added a new feature - Software Engineered For Excellent Performance 

One would expect Apple would show more love with their own music production software considering the hardware is their design.

 

[des99]

They changed the relationship of the Pro/Max chips with the M3. This has nothing to do with Logic. There is a bigger difference between the Pro and Max chips on M3 than the earlier generations. The Max chip is great, and is the one you want for demanding work on M3.

[Richard Herczeg]

5 minutes ago, des99 said:

This has nothing to do with Logic.

What I was referring to has everything to do with how Logic Pro is functioning.

The video I've linked explains that Logic Pro isn't making good use of the Efficiency Cores in any Silicon chip. The newer chips seem to favour Efficiency Cores over Performance Cores.

I'm suggesting that Logic Pro is terribly inefficient. Pun intended.

[yarns]

this guy is hammering his systems though and tests logic after thrashing it for an hour at max capacity before he even runs a test!!!

this guy iz mad, or really rich hahaha.

i'm sure it doesn't matter but just funny to note how hard and long he's stressing these systems, not really a idealistic test environment from a scientific point of view. i think. :shrug:

[yarns]

he's good at math too hahaha

but seriously, how can reaper have better developers than Apple, it's not really acceptable i reckon. well it definately shouldn't be. but then again... if they're not supposed to be using the efficiency cores like they're for the system or something.. then reaper is just cheating and Apple should re-think their design decisions

[yarns]

maybe they're preppin for AI to utilize everyones spare cores while we're not lookin...

to mine bitcoin!

🫠😱🥴😭🤯👻👀🗯️

💰🔥🖥️🔥🌐🔗🔥🖥️🔥

[wonshu]

@yarns

Without dissing the Logic developers AT ALL I have to say that the Reaper developers are pretty much up there with the best of them!

Edited November 29, 2023 by wonshu

[des99]

8 hours ago, Richard Herczeg said:

What I was referring to has everything to do with how Logic Pro is functioning.

The video I've linked explains that Logic Pro isn't making good use of the Efficiency Cores in any Silicon chip. The newer chips seem to favour Efficiency Cores over Performance Cores.

I'm suggesting that Logic Pro is terribly inefficient. Pun intended.

Ok thanks, it's helpful to actually mention the point you want to make in your post to make as we can't all watch an indeterminate length video to try and find it in there somewhere.

Logic has traditionally been one of the most efficient DAWs out there in terms of the audio engine, and the first to adapt to Apple's hardware and macOS features. I don't really know what the issue is, but if there is some issue on the M3 chips that isn't solved by telling Logic what cores you want to use (it's settings by default will favour using the performance cores), then we might see further optimisations in this regard in future versions possibly.

[yarns]

18 hours ago, wonshu said:

@yarns

Without dissing the Logic developers AT ALL I have to say that the Reaper developers are pretty much up there with the best of them!

I know it's a bit harsh but I was just making conversation. I actually have no idea why some of those DAWs wern't using the efficiency cores as I'm not into that kind of thing at the moment but that's why I'd be weary of apps that reap every resource at their disposal. There's gotta be a reason (good or not) that Apple keeps altering the ratio of cores. I have no idea but I'm interested now... I'll look into it 🤔🐈💣🐈‍⬛🧨

[des99]

10 minutes ago, yarns said:

There's gotta be a reason (good or not) that Apple keeps altering the ratio of cores. I have no idea but I'm interested now... I'll look into it

Yes - for the M1, the Pro and the Max chip essentially had the same CPU setup in terms of core layout etc. The M1 Max simply added the potential for more RAM, more GPUs, extra neural cores etc.

They've changed this setup for the current M3 generation, probably both to a) differentiate the Max and the Pro a bit more and I suspect b) feedback from how the machines were typically being used suggested that making the "Pro" a little more consumer-focused meant that actually people would find more e-cores useful for regular work, as the p-cores were likely only being very lightly used - especially as all the cores continued to get more powerful, and the e-cores are plenty capable of doing the bulk of the users' tasks.

For the people that heavily exploit the CPU/GPUs (for example, DAW users), those people benefit from the best p-core setup, and this is now what the Max machine is intended to be for. There is now correspondingly a much bigger difference in processing power between the Pro and Max on M3.

Out of interest, for the last few weeks I've changed up my CPU monitor to display all the core activity. My M1 Pro has eight p-cores and two e-cores, and for generally computer, the e-cores get hammered quite a bit, and the p-cores only kick when doing something that really ramps up multiprocessing.

So it's a product choice that Apple made that they presumably feel provides the best range in performance for their target markets. If you want processor grunt for heavy tasks, the Pro is not as good as it used to be, relative to its position in the range.

[mparker]

On 11/29/2023 at 12:01 PM, des99 said:

They changed the relationship of the Pro/Max chips with the M3. This has nothing to do with Logic. There is a bigger difference between the Pro and Max chips on M3 than the earlier generations. The Max chip is great, and is the one you want for demanding work on M3.

That’s all well and good if you can afford it! Many users can’t and also the iMac, which I prefer has no Max chip option. So semi-pro users are petty well up the creek or we use another DAW. Pretty shoddy imho! I wouldn’t mind if there was some clear rathole available to explain to users, but I spoke to a Logic Pro techie at Apple yesterday and he simply spouted banal generalizations and was pretty dismissive. World class service, 😂! 

[des99]

38 minutes ago, mparker said:

Many users can’t and also the iMac, which I prefer has no Max chip option.

The iMac doesn't have a Pro option either, it's the base level Mx-chip so it's always going to be performance constrained compared to the more powerful machines. Apple don't seem to want the iMac to be a pro-level machine anymore, they want it to be a more consumer/family/business machine it seems.

I don't know what the Geekbench scores are, but I'd be surprised if an M3 iMac would be slower than an M1/M2 iMac though...

38 minutes ago, mparker said:

That’s all well and good if you can afford it!

Which has always (alas) been the case with Apple...

[Dewdman42]

I think personally, Apple, Presonus, Imagine-line and ableton will be looking at their products to see how they can provide an option for their DAW's to use all the cores.  its ridiculous to leave efficiency cores sitting idle and the DAW underperforming.  Kudos to Steinberg and Reaper for realizing that already, I'm sure the others will be paying attention and providing options to do so in the future in order to keep up with Reaper and Steinberg.  It's a significant drop in performance  while also making the M2 and M3 Pro's with less performance cores!

For now, I would say for audio production a used M1Pro would be a great choice, presuming it has enough RAM for you.  For me, if I were going to get an ARM mac, I would want larger RAM which pretty much means I would need a Max anyway but not everyone is like that for sure and its a heavy premium to have to pay when I really don't need all the GPU processing.

M1Pro is great value though for many people though I would say. 

 

[Mark Horoocks]

There's a great video here which shows the weird performance issues. I'm fine on the bas model M3, but it is frustrating when Apple's own door is outperformed on their own hardware. I just wish I knew what the (pardon the pun!) logic behind it is.

 

[des99]

Could you sum up the issue for this forum? Was Logic set to use the e-cores in the audio preferences?

[Dewdman42]

Does logic pref have a way to do that?  I hear someone suggest you change the number of threads from automatic to a larger explicit number might do it

[des99]

1 hour ago, Dewdman42 said:

Does logic pref have a way to do that?

Yes, hence the question...

My machine has 8 p-cores, and 2 e-cores. If I set Logic to use 8 (8 p-cores) (which is my default), it won't assign work to the e-cores. (I leave them to the system, as recommended, for best fluid overall system performance).

If I select 10 (8 p-cores), it *also* assigns work to the extra 2 e-cores as well (with whatever scheduling algorithm it uses.)

So, if someone is using Logic on a machine with (hypothetically) 2 p-cores, and 8 e-cores, and Logic by default on those machines only uses the p-cores (I do not know whether Logic defaults like this on those machines), then you're only using a small amount of the available CPU.

I haven't watched a bunch of videos, but depending on what was said, perhaps the video chap doesn't know how this works, left Logic on the default, and saw a decreased performance over earlier machines that had more p-cores. I can't really comment on that, as I haven't watched the video. If someone wants to actually let us know the key concept in the video they are asking about, and whether it's a different issue, I might be able to comment further.

Please, when asking about a video someone has posted, *also* include a sentence or two describing the issue. It's painful to have to hunt randomly for the context in a 10+ minute video when a sentence would let us know in a second.

[Dewdman42]

2 minutes ago, des99 said:

Yes, hence the question...

My machine has 8 p-cores, and 2 e-cores. If I set Logic to use 8 (8 p-cores) (which is my default), it won't assign work to the e-cores. (I leave them to the system, as recommended, for best fluid overall system performance).

If I select 10 (8 p-cores), it *also* assigns work to the extra 2 e-cores as well (with whatever scheduling algorithm it uses.)

So, if someone is using Logic on a machine with (hypothetically) 2 p-cores, and 8 e-cores, and Logic by default on those machines only uses the p-cores (I do not know whether Logic defaults like this on those machines), then you're only using a small amount of the available CPU.

I haven't watched a bunch of videos, but depending on what was said, perhaps the video chap doesn't know how this works, left Logic on the default, and saw a decreased performance over earlier machines that had more p-cores. I can't really comment on that, as I haven't watched the video.

Someone needs to point that out to the kid who made that comparison video

[des99]

The problem with YouTube videos is someone makes a video about something they don't understand (instead of actually investigating it), then other people treat those videos as canonical, and now think the software has a problem, and start blaming Apple for their terrible inefficiency and poor coding, rather than checking out the behaviour, reading the manuals, and determining for themselves whether there is an *actual* issue.

So - for the above posters commenting about this video - *is* there an *actual* issue, or was it simply that the person in that YT video didn't understand how to set up his audio prefs for his machine?

If there is a *real* issue, we can look into it. Otherwise - don't believe everything on YouTube is coming from informed sources...

[Dewdman42]

The video is still interesting as it compares half a dozen daws and attempts to compare the three m chips on each daw.  I found it interesting and it’s well done but he should have set that setting in logicpro and obviously didn’t know about it.  He should really redo the video now.

peoppe post non video things like this all over the internet the problem isn’t you tube per say its the internet and social media and people being able to easily publish information in any form to a wide audience.

what I dislike about these videos is that you have to sit through 20 minutes of dialog in order to get what could have been presented in a blog post that takes 2-3 mins to read.  But videos are sexier and in some cases generate more revenue so everyone is out there trying to be Oliver stone 

[Dewdman42]

I just watched the video again the section about logicpro and he makes a big deal about how he did actually set the number of threads to the max setting and still the results were not good for logicpro.

 

[synth_hero]

Quote

Please, when asking about a video someone has posted, *also* include a sentence or two describing the issue. It's painful to have to hunt randomly for the context in a 10+ minute video when a sentence would let us know in a second.

This is a perennial issue on social media and internet discussion boards. I saw the video in question but needed time to investigate before commenting. 

The video poster did not mention anything about configuring the number of performance cores.

[des99]

3 minutes ago, Dewdman42 said:

I just watched the video again the section about logicpro and he makes a big deal about how he did actually set the number of threads to the max setting and still the results were not good for logicpro.

Ok - what does "not good" mean? Is there actual data for how much CPU an e-core gets on M1/M2 and M3, and how they relate to p-core performance, so people thinking about getting an M3 non-pro machine can estimate the performance they get?

I seem to recall that someone said the M3 e-cores are actually more performance than the M1 p-cores, but I'm not sure of the source or accuracy of that...

[Dewdman42]

At this point I recommend you watch the video.

logicpro is not using the efficiency cores even when configured to do so.  Honestly the results are pretty dissapointing.  Easy fix for Apple to catch back up to reaper and cubase in this regard but as of now it’s 30% worse.

[Dewdman42]

9 minutes ago, synth_hero said:

This is a perennial issue on social media and internet discussion boards. I saw the video in question but needed time to investigate before commenting. 

The video poster did not mention anything about configuring the number of performance cores.

Yes he actually did, I just watched it.  Unless he updated the video it was always there that he did that

[synth_hero]

7 minutes ago, Dewdman42 said:

Yes he actually did, I just watched it.  Unless he updated the video it was always there that he did that

A timestamp would help, thanks. 🙂

[Dewdman42]

The fundamental issue is that more than half of the daws tested here are not using the efficiency cores for processing tracks in the daw.

there may or may not be some fundamental theoretical reason why Apple chooses not to use efficiency cores in logicpro, but the bottom line in the test is 30% less tracks.

the test then exposes a further point that because the series from m1 to m3 has changed over time to have less performance cores in each iteration, that has made daw performance worse with each iteration with the exception that of reaper and cubase.

of course you can buy a max chip to get more performance cores.

all daws should utilize the efficiency cores unless there is some specific reason why that should not be done 

5 minutes ago, synth_hero said:

A timestamp would help, thanks. 🙂

You can find it as easy as i did 

Edited December 3, 2023 by Dewdman42

[Dewdman42]

So waxing philosophical about this for a moment.  I haven't followed the M chip advancements that closely because I'm still using my trusty ol' maxed out 2010 cheese grater for the time being...it will be at least two more years before I will upgrade to Apple Silicon for my DAW.  but anyway, this conversation has pulled me into it a bit.

Why less performance cores?

First, the question is why is M2Pro using less performance cores compared to M1Pro and again the M3 Pro less then the M2Pro?  Answer, in order to have more efficiency cores.

That one is easy, Apple is trying to differentiate their audiences, an awful lot of people could probably get by with nothing but efficiency cores, but maybe more of them.  So M2Pro has more efficiency cores (and less performance cores) and likewise M3 Pro even more so.

Efficiency cores use 1/10 the power and battery apparently compared to performance cores.  But provide "enough" computing power to most things that people are doing on their computer, especially if they have more cores.  

Battery life means almost nothing to me as a DAW user, but we are a small percentage of their user base I think.

Why not use efficiency cores in the DAW?

Easy answer, it should.  Of course they are less powerful than the performance cores, so what are the implications of using efficiency cores for DAW use?  Steinberg has apparently found a way to do it and Reaper too, but that doesn't mean it might not be complicated in terms of multi-threading algorithms and efficiency cores might require larger buffer settings and other problematic things.  Thus we see in the case of some of these DAW's they are literally ignoring the efficiency cores. 

I get why apple is doing what they are doing with the M chip line, attempting to make the Pro line just a little bit more powerful for common every day tasks that the vast majority of people are doing, by having more efficiency cores...and will fundamentally use less battery life for what they are doing without having to tap into performance cores very often.  However, for DAW use, this quickly starts to make make DAW users more candidates for the max line due to this change of core allocation rolling forward, as DAW processing is a task that needs power cores.  but on the other hand, Steinberg uses the efficiency cores somehow, but the question there is, how does steinberg use the efficiency cores without ending up with longer buffer  requirements and other things to be able to use the efficiency cores for DAW?

The fact that LogicPro's preferences provides a way to specify to use all the cores, yet still doesn't seem to use the efficiency cores..is interesting.  That could simply be a bug that needs fixing.  But the way it's labeled is a bit misleading.  In the past that preference was labeled as being the "thread" count, not the core count.  Threads are typically allocated to cores by the OS not by the app.  So it's possible that LogicPro is leaving it to the os to use the efficiency cores or not, and for whatever the reason Steinberg is able to trick MacOS into using the efficiency cores and LogicPro is not.  Something like that...  just thinking out loud.  Only Reaper and cubase are using the efficiency cores and its not clear to me if there might not be some other unintended consequences of using the efficiency cores in parallel with the performance cores for processing the tracks full of plugins in a DAW project.  Its entirely possible that Apple is actually doing it the right way and that the renegades Reaper and Cubase might initially look to be using the efficiency cores better...but may end up with weird behavior in some way that hasn't been exposed yet from this test at least.

bottom line for daw users though, is that newer M Pro chips will be less capable in some daws compared to the older ones...and daw users should heavily consider making the investment in the max version in order to have more performance cores.

 

[synth_hero]

If we could reproduce the video poster’s tests on our individual setups then we could go from there. 

[Dewdman42]

If I had an Apple silicon mac I would already be doing it, but I don't.  Hope some of you will.  It would be good to figure out if its possible to goose LogicPro into using the efficiency cores in addition to the performance cores.

The interesting thing is, the purpose of efficiency cores is to use less battery with less capable cores whenever possible.  Theoretically the efficiency cores should be the ones always loaded up and the performance cores should only come on sometimes when that power is needed.  But how an what determines when to jump to performance cores or not..is beyond my own knowledge, but we just see on the meters from several DAW's that when processing around 80 tracks, the performance cores are loaded up and the efficiency cores are dead quiet. 

 

1 hour ago, des99 said:

Ok - what does "not good" mean? Is there actual data for how much CPU an e-core gets on M1/M2 and M3, and how they relate to p-core performance, so people thinking about getting an M3 non-pro machine can estimate the performance they get?

I seem to recall that someone said the M3 e-cores are actually more performance than the M1 p-cores, but I'm not sure of the source or accuracy of that...

Des, please watch the video if you want to discuss it.  thanks