Sascha Franck

No worries about that from my side at all. At least not for the stuff I'm doing atm. In general, I could keep using my MP until the end of times and wouldn't miss much (others than it not being able to run any up to date software). The only thing I sometimes wish for is more single-thread "oomph" - and the M3 should deliver that in spades, at least compared to this aging Xeon machine. And it wasn't ever much about that anyway. But I'm having quite some concerns about QA in Logic land (admittedly nothing new, at least not on my side). We shall see...

Fwiw, I ordered a MBA, so I can see for myself what's going on with the E-cores.

I did exactly nothing. As soon as a track is not in live mode anymore (hence not selected and/or record enabled), it automatically becomes "distributable". Since ages already. If you want, I can post a video. Or a GIF. Whatever.

I have absolutely no idea. That is simply not true. I tried with DOZENS of instrument and audio tracks - Logic splits them all onto two threads in playback mode (which is the relevant situation for the E-cores being able to work properly). Every single one of them without even just one exception. I have a very good idea what it looks like because I've seen it in person. And I've said so multiple times already.

Very unlikely. And even if it was like that, Logic seems incapable of doing so (see Reaper/Cubase). From all that I can see, Logic can split individual channel loads onto different threads in playback mode just fine. Nothing else is needed to "beam" a task from a P-core to (multiple, if needed) E-cores. And for the umpteenth time: With the M1 CPUs that just seems to work fine.

I'm not sure what this is supposed to demonstrate. Doesn't seem like E-core utilisation has much to do with Logic here.

Sorry but again: This is completely irrelevant. This thread is about: - Logic not supporting E-cores on newer machines anymore when it has before. - The competition supporting E-cores and making Logic look bad in comparison.

I wouldn't call that a "fair bit" - and it's very likely (so likely I'd pretty much bet on it) they're not used within Logic. I mean, that screenshot is taken pretty much the moment before Logic is crapping out, so the E-cores are defenitely not helping it in balancing the load. And then, as said numerous times already: It's totally different in Reaper and Cubase. No matter how you put it, 105/104 vs. 64 tracks on the same machine is a most significant difference. So, Logic not using the E-cores directly translates to its overall performance.

Again: This thread is not about what I need. And I defenitely don't need advice.

Sorry to say so, but all that is pretty much irrelevant, @vondersulzburg. The relevant thing is: Logic was using the E-cores fine on the M1 Macs I tried with (a Mini and a MBA), load distribution was pretty even. This can as well be seen in some YT videos. Now, with the event of the M2/M3 CPUs, the E-cores are almost not in use at all - which results in roughly 40% loss of overall computational power. That's a whole lot. And fwiw, this thread isn't about the CPU power I need right now. Or about what I may need next month. Or next year. I was just about to buy a new machine when I stumbled over that video, so I thought it'd be a good idea to inform myself whether it's true. Nobody needs to tell me what I should possibly buy instead, that the CPU juice would be fine anyway and what not. Hence, this thread is entirely about Logic not utilizing the E-cores anymore. Which, in my book, is an absolutely inacceptable thing (especially given all of Apples marketing blurb a la "more Pro than ever", "get more done in less time" and what not - also including the praise the efficiency of Logic always gets). And btw, in my book Logic isn't exactly cheap. Simply because you need to buy a Mac to run it. And you need to do so more often than the hardware would possibly require (I'm only purchasing a new machine now because the lack of support for my Mac Pro is really starting to show). Anyhow, even that isn't relevant. Logic was doing fine with M1 CPUs, so it should as well be doing fine with M3 CPUs. But it isn't. It's even way less efficient than the competition - which is a first.

Well, doesn't need to be realtime/live. It's fine in case they're available for playback purposes. Really, just as things were with the M1 CPUs.

Correct, typically it's two. Which should be fine to distribute some load from a P-core to two E-cores. I would pretty much place almost any bet it's done by Logic - not by the plugin. Why? Because the load is always distributed evenly. Can't imagine plugins such as, say, Helix Native, which are highly modular in nature, are always recalculating things (especially as in general, the native version is sharing its codebase with the hardware). Yes, that is my dumb user/layman understanding, but it's simply the most plausible thing to happen, as you're not depending on the plugin maker - who might as well not care and bring everything to a grinding halt. But once the host distributes the load, it'll just do it with each and every plugin (and Logic does just that, it's working with every single plugin I ever stumbled across). I'd even go as far as to assume it's not even happening on a per-plugin base but on a per track/channel base. Regardless of what I'm doing in Logic, the load is always distributed to two playback threads evenly (minus things such as Alchemy which seems to take that to the next level as in being distributed to 3 threads - but that might only work because it's more or less part of Logic). However, this is working with each and every plugin and it's not new in Logic, either. And this very functionality would as well be absolutely perfect to distribute heavy P-core loads to 2 E-cores. I think it's got a lot to do with E-core utilisation because IMO proper multithread handling is a pre-requisite for it to work. And when you think about it, it makes a whole lot of sense, too. However, that obviously doesn't explain why Logic fails so miserably with the M3 Macs. And it's also why I was incredibly astonished to stumble over it. Because: - Other hosts are doing just fine (so it's not like the E-cores would be "force reserved for system tasks" or whatever some people already came up with). - The latest Logic incarnations are actually doing amazingly well in terms of multithreading. - Things worked just fine with the M1 CPUs (I've seen it personally). Only in "live" mode and on serial signal paths. As soon as a channel is set to playback only, its load can be distributed to multiple cores. On live threads that wouldn't work as the latency would rise. As said, I think these two are directily connected. Sure, doesn't explain why Logic fails at E-core utilisation, but it's still very important to know that Logic is very well able to split CPU loads of single channels to multiple threads - which pretty much renders analogika's argument void.

So, Reaper and Cubase - how do they manage?

I doubt that. It's the same with each and every plugin - even with Diva and "Multicore" switched off. I have not stumbled across a single plugin that wouldn't spread its load over multiple threads in playback mode.

Maybe, maybe not - but the generally way worse performance doesn't speak in favour of that, especially when you compare to Reaper or Cubase. I hate saying so, but that very complaining has led to some nice improvements (as you may or may not know, I've been betatesting for Emagic back in the days). And btw: Logic can do that already. This is a single instance of Alchemy playing, nothing else. So that would explain why things were working just fine utilizing the E-cores on M1 machines.