Music Production - Poor performance of certain DAWs on M3Pro

[MikB]

Maybe you should get off the forums once in a while and make some music, as I do.

Your assumptions continue. How do I know things? I test and compare notes. You reverting to ad hominem only shows you got nothing.
Ask some real questions instead.

 

[MikB]

The problem with utilizing E-cores in the DAW is that it's sufficient that one single core is overloaded to cause audio buffer under-runs resulting in crackling audio. Which has been suggested already by several. Of course, it might be macOS is smart enough to handle that, but I suspect the DAW isn't giving the OS the required info about how much processing it might need, so I'd start with the assumption the DAW user can have no control over this.

Obviously, as DAWs typically will process any specific track on one core, spreading out the load would be beneficial as long as no track overloads an E-core, but remember with track count going up you will increase the amount of tracks that are processed on a single core plus macOS will also make use of the E-cores.

Different DAWs will spread out work differently when recording for instance, potentially worsening the situation, but there's certainly nothing wrong testing your Mac and see if it works in your case. With the M3 offerings with so few P-cores maybe it's even advisable to try, if you already got one.

Personally I've noticed a very even distribution when using external wrapping software that does utilize also e-cores. But I don't use wrappers all the time and I wouldn't rely on this function for purchase decisions, even if my DAW seemed fine with it.

 

[MikB]

I can't say for Apple Silicon, as I don't have it yet. However, this situation is somewhat similar to my Intel Mac, which has 8 hardware cores, but can run 16 threads using hyperthreading. And these additional 8 cores are disabled in Logic by default as well. Enabling them has no any bad consequences such as lower stability or higher latency. This is just a free performance, disabled by default for unknown reason.

This is not directly comparable points as the CPUs are so different. Apple Silicon may have to abide same laws of physics, but the integration and deep knowledge Apple has of the Mx design has changed the game. Mx is a completely other kind of experience.
That said, I don't think anyone interested should refrain from trying to enable E-cores for music. Just don't expect this to always work, or to be fully controllable, and you'll be fine I do think. So far, I haven't really even needed to get deep into this. I don't have a CPU power shortage and I'm on M1. What's my incentive?

 

[MikB]

It's not a comparison between Mx chips, it's a comparison of different DAWs and how they utilize those chips. The point being if the M1 Pro has more performance cores than the M3 Pro, in this particular instance certain DAWs would perform better with the M1 Pro. At least in this type of scenario. That and a reduction in memory bandwidth seems to play an important role. This is an important point for music producers looking to upgrade from an M1 Pro to an M3 Pro for example.

These are indeed important points to consider, but memory bandwidth is possibly a red herring. Those slower speeds are still very fast.
It's somewhat bewildering you both try to put a light on P-core counts changing in various generations of Mx and at the same time you're implying trying to use E-cores for audio could be a solution. Are you actually trying to defend Apple M3 offerings?
Would I consider the Max because of higher P-core count? Yes.

 

[ProgRocker]

I appreciate the clarification that you're not comparing the M1 Pro and M3 Pro directly, but rather how they perform with different DAWs. It's definitely an important consideration for music producers who rely heavily on specific software.

Yes, it's an important consideration in choosing both an appropriate CPU and DAW for your specific application as we all have different needs. Each DAW and CPU combination brings something different.

 

[3SQ Machine]

I've been testing my same logic projects on M1 Pro 16", M2 Pro 14" (binned), and M3 Pro 16".

Here's the strange part -- less crackles/pops on M3 Pro and definitely less % CPU on same projects compared to M1 Pro... BUT, graphics are less "smooth" than M1 Pro. Scroll in play in particular is more choppy where the tracks stutter a bit on their way across the screen. It's annoying since I use this feature for playback and it visually looks like a downgrade on both the M2 Pro and the M3 Pro.

I've tried to troubleshoot but not much guidance since it's not system overloads or even high CPU usage to blame. Buffer sizes, etc., make no difference. I can only guess that 2 less performance cores somehow gimped the graphics. I've tried all the usual "optimizations" but none make it as visually smooth as the M1 Pro.

Last step will be a potential nowhere call to Apple Support since the machines are covered and I'm in the return period. I got some great deals on these new laptops and was hoping this new setup could retire the M1 Pro. Now I'm not so sure.

 

[Zest28]

I appreciate the clarification that you're not comparing the M1 Pro and M3 Pro directly, but rather how they perform with different DAWs. It's definitely an important consideration for music producers who rely heavily on specific software.

Yeah man, makes sense to drop Logic because some people on MacRumors and YouTube are spreading misinformation that Logic cannot use efficiency cores.

 

[ProgRocker]

I've been testing my same logic projects on M1 Pro 16", M2 Pro 14" (binned), and M3 Pro 16".

Here's the strange part -- less crackles/pops on M3 Pro and definitely less % CPU on same projects compared to M1 Pro... BUT, graphics are less "smooth" than M1 Pro. Scroll in play in particular is more choppy where the tracks stutter a bit on their way across the screen. It's annoying since I use this feature for playback and it visually looks like a downgrade on both the M2 Pro and the M3 Pro.

I've tried to troubleshoot but not much guidance since it's not system overloads or even high CPU usage to blame. Buffer sizes, etc., make no difference. I can only guess that 2 less performance cores somehow gimped the graphics. I've tried all the usual "optimizations" but none make it as visually smooth as the M1 Pro.

Last step will be a potential nowhere call to Apple Support since the machines are covered and I'm in the return period. I got some great deals on these new laptops and was hoping this new setup could retire the M1 Pro. Now I'm not so sure.

Interesting, thanks for sharing.

 

[3SQ Machine]

Interesting, thanks for sharing.

Quick update -- called Apple Support and got the Logic department. Seemed well-versed and shared screen to show stuttering. Made a few adjustments, reset plist, etc.--no luck. Basically, it looked like a 60 hertz refresh rate despite promotion being on.

According to him, likely a unit-specific problem with the display. I of course mentioned the loss of P cores to which he said nothing. Suggested I return it instead of Apple engineering taking the computer for weeks and blowing past the return deadline. I took the suggestion as my gut said this was an M3 Pro problem and not a unit problem.

And I think I'm right. So while bummed because it was a steal of a deal for the 36gb M3 Pro unit ($600 off), the store amazingly was in the middle of shipping off my 14" M3 Pro (unbinned) that I returned days earlier to go with the 16". They "sold" it back to me for the same sale price (after I literally chased down the truck it had been loaded on).

Doing a very quick project on the 14" M3 Pro -- stuttering not as bad in scroll-in-play on the smaller screen (more livable) but still, M1 Pro smooth as butter. Is it the loss of 2 P cores and who-knows-what threading priority for the display refresh rate? Who knows, but support mentioned something about reports of "lag" on the M3 pro with Final Cut Pro app also.

Anyway, I'm done--definitely not trading "up" my 16" M1 Pro for the 16" M3 Pro. Maybe someone who can't A/B test with the M1 Pro would never notice. It's certainly more livable on the 14"--but I hadn't really intended this as my main unit, just a road unit. It's fine for that.

 

[Basic75]

Perhaps it's the different cache layout?

The M1 Pro has two clusters of 4 P-cores, each cluster sharing a 12MB L2 cache. In the M3 Pro all 6 P-cores are in a single cluster with 16MB of shared L2.

If the software in question is very cache sensitive that might be a reason to avoid the E-cores even though they are not super slow, each E-core cluster, 2 cores in the M1 Pro and 6 cores in the M3 Pro, only has 4MB of shared L2.

Not sure how to test if the cache is the problem.

 

[3SQ Machine]

Perhaps it's the different cache layout?

Not sure how to test if the cache is the problem.

This is a great point. No reason for this machine--on paper--to struggle with the refresh rate on an (Apple) pro app--except for some kind of architectural bottleneck. I'm going to test out logic some more at the Apple store on the M3 Pro 16" and see if I can replicate the problem. There's really nothing out there talking about this issue on the M3 Pro, strangely enough. I would think a lot of pro users, who are even more particular than I am, would pick up on this, unless the M1 Pro/Max crowd is going for the M3 Max or sitting out this upgrade cycle.

 

[3SQ Machine]

Update--went to Apple store to test the logic issue on M3 Max and another M3 Pro unit they had. Same stuttering/refresh rate issue on all units, even the Max. I talked to the Apple Store DAW experts who were complaining about it--it's something logic related. It is affecting even older units although not my M1 Pro. The good news--it's on their radar screen and may be fixed someday (or not). The bad news? I had taken Apple's advice days earlier to return my 36gb M3 Pro that I got for an unheard of $600 off (and can't get now). Lesson learned to FULLY investigate a problem before returning.

 

[ProgRocker]

The good news--it's on their radar screen and may be fixed someday (or not).

Thanks for the update. Glad to hear it's on their radar, that lends hope they may address this in the future.
Did they happen mention any other DAWs, or is this just something with Logic ?

 

[3SQ Machine]

Thanks for the update. Glad to hear it's on their radar, that lends hope they may address this in the future.
Did they happen mention any other DAWs, or is this just something with Logic ?

No, other DAWs didn't come up. It was easier to test logic because the demos all have it installed.

 

[3SQ Machine]

One quick interesting finding on the "choppiness"--for some reason, logic is spiking the GPU usage to just shy of max when there are a lot of tracks it is trying to scroll. These spikes correspond exactly with the choppiness. When the usage goes down (as monitored in activity monitor) it is smooth again.

My M1 Pro with two less GPU cores than the M3 Pro does not suffer from these spikes, and is also much more smooth. I haven't run the same GPU monitoring on the M2 Pro 10 core yet so I can't determine if it is something architectural with the M3 GPU.

Also, as an aside, Best Buy was willing to "sell me back" the 36gb unit I returned for the same $600 off price so now I'm determined to see if I can get this resolved. At least I know it's not a CPU issue.

 

[3SQ Machine]

Sent log files and a screen recording to Apple Support. Logic guys stumped (since it clearly makes no sense).

Issue also exists on 10-core binned M2 Pro. Looks like without a resolution there will be no "upgrade" for my music production unit beyond the M1 series. Will report back if any fix ensues.

 

[3SQ Machine]

Brought my projects to the Apple store where they let me load them up to test. Every single unit had the same issue--logic taxing the GPU hard to cause choppiness. Clearly an optimization problem. Even the M2 Max Studio and the M3 Max were less smooth than the M1 Pro. Utterly and completely unacceptable.

I think people using live loops or not doing scroll-in-play would never notice a performance issue. Even if zoomed out a bit it calms down--it just so happens to be that MY way of viewing the timeline at MY preferred zoom level is gimped in the post-M1 Pro versions.

Thanks for letting me vent on this thread. I've been through so many new laptops the last couple months trying to get the "right" one to replace the M1 Pro for music production and Apple has failed me.

 

[3SQ Machine]

Sitting here typing this on a binned M3 Max 16" I found at MicroCenter open box for $2677. While I'm impressed with my ability to get a discount on expensive Apple products I remain less impressed with the products themselves.

So--the interesting news--this is the "smoothest" of all the units I brought home to test. Side-by-side with the M1 Pro I would call them equal. But, calling a new, more expensive machine equal to the first gen in UI smoothness is no compliment. The choppiness is reduced, but it is no better than the M1 Pro.

Of course, I'd be a fool to say this machine isn't an upgrade. Uh, it's so ***** fast I laugh at how instantly everything loads. There's no point in having macos animations because they are all like 10x speed on their transitions. Plus, this has 2 more blazing-fast performance cores than my M1 Pro so there's no doubt this would improve my workflow if I keep it.

Oddly, my test projects in the Apple stores on the demo units (I tried out two top M3 Max's at the stores with 48gb ram) didn't show the same smoothness. But it could be because the demos are all running 10.7.9 (I checked) and maybe there are some other performance limits on the demos that hurt results.

I'm also going to test out a base model 14" M1 Pro that I brought home to see if that has the same level of smoothness and will report back.

I should start a career as a youtube reviewer after all this.

 

[3SQ Machine]

All tests done. Base M1 Pro 14" with binned chip no better. Glad I didn't waste anymore time on that.

Conclusions? After testing the same LP projects on gawd knows how many machines at this point, here are my takeaways:

• Losing a performance core or two on M3 series does not hurt my projects in terms of raw processing power of tracks & plugins-- there are less clicks & pops and more headroom on M3 Pro v. M1 Pro (again, for MY projects, so YMMV).
• For the smoothest Logic UI with scroll-in-play it seems 8 performance cores + 16 GPU cores minimum. The unbinned M3 Pro with 6 performance cores and 18 GPU cores was not smooth, so I can only guess that the performance cores are also playing some role with graphics optimization. The binned M2 Pro with 6P + 14GPU was also not smooth.
• M1 Pro is superior to all others except M3 Max (binned) in Logic UI smoothness & graphics. If you like to edit using scroll-in-play like I do, you will dislike the M3 Pro and anything less than a binned M3 Max.
• Binned M3 Max is a beast and a clear successor to the M1 Pro, but disappointingly is not any smoother in graphics/UI--likely limited by software optimization. It's still an expensive proposition and may not be worth the money.
• A 12 core M2 Pro/M2 Max will likely do the trick--I suspect it would but never tested it. Oddly, the M2 Max Studio in the Apple Store did not handle my projects with any more smoothness. It was also connected to an Apple Display at who knows what resolution/refresh rate. Bottom line--more tests needed.
• Best value still is unbinned M1/M2 Pro for logic if interested in the smoothest UI experience + power.

Now maybe thanks to my opening a case with Apple on the M3 Pro there will be some update made, since it honestly makes no sense that logic would tax the GPU on any of these machines. This is not 3D rendering to just scroll some tracks in different colors with some text across the screen. However, until they do--if they ever do--best bet is to aim for the right specs to get the best experience.

 

[ProgRocker]

(again, for MY projects, so YMMV).


For the sake of comparison, what generally constitutes one of your projects ?

 

[3SQ Machine]

I don't know why, but the quote feature is all buggy so I can't quote. Long story short -- my projects tend to demand high single-core performance, not because this is ideal but because I've been more/less efficient at times in sending effects to other buses. I've got too many live projects to go back and correct this, but fortunately most Apple silicon has been excellent despite my inefficiency. I do not use large sample libraries, but just a lot of effects/synths/midi/plug-ins--so I don't need a lot of ram. Generally, that captures my workflow.

Also, if no one is zooming in and using scroll-in-play, they probably would be just fine. It's just this one specific way I edit where I encounter a huge downgrade on some of the newer Pros.

 

[3SQ Machine]

Quick--bizarre--update. Bought an Apple refurb 14" M1 Pro with the full 10 core CPU/16 core GPU. Same specs as my 16" M1 Pro, just smaller. Tested the project--same problem! I'm stumped.

Starting to wonder if there is something "special" about my 16" M1 Pro that makes it run smoother with less CPU/GPU %. Geekbench/Cinebench on both machines show identical CPU/GPU scores and there is no evidence of any throttling on the 14". I've done clean installs as a new user (no migration) with a fresh logic install/ no user plugins or prefs on brand new machines and I'm not close to understanding this mystery. I've checked for any other settings that could be different but there is nothing obvious. M1 Pro 14" should equal M1 Pro 16", no?

I will say that the 14" shows more processes and threads in the CPU under activity monitor than the 16". Whether that will settle down over a few days as the machine finishes indexing and other setup background tasks, who knows.

Was thinking that maybe logic for some reason has to work harder on these 14" displays because of the resolution and scaling, but the 16" M3 Pro had the same performance issue with GPU spikes/cpu usage--but there are also 2 less performance cores (but more GPU cores), so not an exact comparison.

At the end of all this, I'm not really clear on understanding why only the M3 Max 16" offers the same logic performance as the 16" M1 Pro. What's weird is that it's not actually "better" with any less GPU/CPU usage when running my projects--it just matches the M1 Pro 16" CPU/GPU usage exactly. Either way, I'm determined at this point to understand this issue so I may grab a refurb 16" M1 Pro and do the same test. If results=the smoothness of my current M1 Pro 16" then I'll have a clue to the "final" answer to this bizarre puzzle--and know more about what machines in the current lineup offer the same performance.

 

[ProgRocker]

Is this just with Logic ? Have you tested other DAWs ?

 

[Chancha]

14” and 16” chassis still have a difference in their power and cooling systems. So even with the identical core config M1 Pros you may still see the two behave differently, if the task happens to stumble upon a limit on the 14” but not on the 16”.

 

[3SQ Machine]

14” and 16” chassis still have a difference in their power and cooling systems. So even with the identical core config M1 Pros you may still see the two behave differently, if the task happens to stumble upon a limit on the 14” but not on the 16”.

Good point and there is a definite difference. Used sudo powermetrics to watch the combined CPU/GPU power consumption on the project. 14" M1 Pro oddly uses 25% more combined CPU/GPU power (in mW) than the 16" in the exact same workflow. I've read that the M1 Max can throttle power consumption in mixed CPU/GPU combined power demands, but the limit I'm hitting on the M1 Pro 14" (5150 mW) seems way below any threshhold where imposed power consumption limits would be hit.

Is this just with Logic ? Have you tested other DAWs ?

Yes, just logic. I don't use anything else. I've read Reaper may actually be better optimized for apple silicon, but without using it I can't comment.