Testing of DAW performance on all 3 Mac M processors.

[Trensharo]

I am actually surprised Logic doesn't use the efficiency cores at all.

Probably has something to do with how they use Rosetta 2 to host x64 plug-ins, etc. Not something you want running on efficiency cores.

They should have just dropped them unless the entire DAW was running in Rosetta 2, like Steinberg did with VST2.

Cubase also has an option to shut down plug-ins when audio isn't going through them, which gives it potential for utilizing additional processing headroom, as well. Not sure if Logic Pro does the same. I never recreate the same project across these two DAWs, so I can't compare 1:1.

[stoopicus]

You should still use them for some things though. In asymmetric multiprocessing development you usually use the lower performance processors to offload non-latency or compute critical processing to. In audio processing you could be using them for a lot of discretionary, noncritical optimizations (pre-rendering, etc). Apple has quite good developer documentation around this, in fact. For example:

https://developer.apple.com/news/?id=vk3m204o

as a high level description with links out to more detailed docs.

This isn't really new; it's been a strategy on non-SMP systems for a long time. Another recent example was the Playstation 3, where the SPU coprocessors were really fast (though limited) and the primary CPU was dog slow. Any game on that platform that was decent at all had to make smart choices about what ran where.

[machinesworking]

You should still use them for some things though. In asymmetric multiprocessing development you usually use the lower performance processors to offload non-latency or compute critical processing to. In audio processing you could be using them for a lot of discretionary, noncritical optimizations (pre-rendering, etc). Apple has quite good developer documentation around this, in fact. For example:

https://developer.apple.com/news/?id=vk3m204o

as a high level description with links out to more detailed docs.

This isn't really new; it's been a strategy on non-SMP systems for a long time. Another recent example was the Playstation 3, where the SPU coprocessors were really fast (though limited) and the primary CPU was dog slow. Any game on that platform that was decent at all had to make smart choices about what ran where.

Blue Cat like I mentioned has some blog posts about the issues they have had with multicore handoff capable audio instruments and fx etc. IMO it has nothing at all to do with Intel AU's in Logic, especially since Reaper and DP both can use that Rosetta 2 tech and do as well, and get in my tests 30 more plugins than Logic.

It is interesting that Logic, owned directly by Apple doesn't use those processors, even when you select them in it's Settings, but I noticed less issues with Logic and plugins that were not "Sonoma ready", at least up until the latest Sonoma updates etc. My guess is the Logic team automagically turn on use of efficiency cores in a future updated when they're certain all issues are gone.

[apoclypse]

Something weird I'm finding with Ableton Live 12 beta is that it does run on the efficiency cores for the most part. Right now my performance cores are pretty much idle while the efficiency cores are getting pinged. Usually it's the opposite.

[machinesworking]

Something weird I'm finding with Ableton Live 12 beta is that it does run on the efficiency cores for the most part. Right now my performance cores are pretty much idle while the efficiency cores are getting pinged. Usually it's the opposite.

That's really goos to hear, I still haven't gotten an invite I should have signed up as soon as they announced it. 11 barely uses efficiency cores, looks like all it uses them for is the basic computer operation etc. I'll do the same test when I get the beta of 12, but I would guess Live would pull off more like 105+ tracks or a 20-24% increase in power over Live 11. That would put it back in normal range VS Reaper and DP, and Logic when it also uses all cores..

[wintoid]

Very curious to hear how Bitwig does in this respect.

[MeanMrMustard]

I'd be curious to hear from the Devs as to why their DAWs don't utilize all cores.

[Jolaff]

Very curious to hear how Bitwig does in this respect.

It doesn't use efficiency core...

[wintoid]

Very curious to hear how Bitwig does in this respect.

It doesn't use efficiency core...

Oh really? Thanks for the info.

[machinesworking]

Very curious to hear how Bitwig does in this respect.

It doesn't use efficiency core...

Oh really? Thanks for the info.

I'm headed to the studio I'll check and edit this post with the results. My guess is it does better than Live, Bitwig in the past did better with Diva than other plugins like Reaktor etc.

[edit]
Scratch my first edit, I had Bitwig at a buffer of 64, at 128 like the rest it gets 96 instances on my system, does not use the efficiency cores etc.

[stoopicus]

Some dev somewhere is facepalming that their CEO said "64 tracks is enough for anybody!"

[machinesworking]

Some dev somewhere is facepalming that their CEO said "64 tracks is enough for anybody!"

To be fair this isn't about running 80 instances of Diva, it's about running dozens of different CPU pigs in one project, or orchestral work with 60 odd tracks etc. Plus I'm using a M2 Mac Studio Ultra, that's the top on Mac OS. The M1 Air here can run only 12 instances of Diva. It still performs almost as well as the 12 core 09 Mac Pro I used before the Studio.

[Trensharo]

Some dev somewhere is facepalming that their CEO said "64 tracks is enough for anybody!"

I think he's right for 85% of the market.

[Dewdman42]

I see two main takeaways from the video

1. Apple has been increasing the number of efficiency cores in the M(n)Pro line with each iteration, at the expense of losing performance cores with each iteration. This probably does make sense for the vast majority of M(n)Pro users who by and large are doing activities for which efficiency cores are plenty powerful enough for them, especially if they have a few more cores then before. And longer battery life for that use case of people. They are optimizing the common use case. DAW users are not common use case, and probably rolling forward we will need to strongly consider the M(n)Max line in order to get not only more memory but also more performance cores.
2. As to why LogicPro is not using the efficiency cores, we can only speculate. I suspect that many long standing multithreading algorithms are now problematic with this architecture and perhaps they simply have to disable efficiency cores for now. On the other hand it's entirely possible that Reaper, Cubase and DP may have unintended consequences with what they are doing that we haven't found out about it. But at least on the surface they appear to achieve much better capability out of the Pro line for DAW use then Apple and the other DAW's mentioned, at least for now.

[Trensharo]

Some dev somewhere is facepalming that their CEO said "64 tracks is enough for anybody!"

To be fair this isn't about running 80 instances of Diva, it's about running dozens of different CPU pigs in one project, or orchestral work with 60 odd tracks etc. Plus I'm using a M2 Mac Studio Ultra, that's the top on Mac OS. The M1 Air here can run only 12 instances of Diva. It still performs almost as well as the 12 core 09 Mac Pro I used before the Studio.

You mean a 2009 Mac Pro?

When M1 hit the market, we were on 12th Gen Intel and 3rd Gen Ryzen. There were 6-8 Core Intel Chips that could outperform that Mac Pro, which was probably hampered more by TDP than the compute potential of the CPU in it (as is the case with most Intel Macs - laptops being most egregious).

PCs didn't have that issue. Put them in a ventilated case with an AIO Cooler of adequate size and they flew. It's why so many people kept those old Intel PC builds in commission for so long. PC laptops also didn't shove i9 CPUs into Ultrabook formfactors, as it makes no sense and just causes the machine to slowly kill itself.

-----

Apple has been increasing the number of efficiency cores in the M(n)Pro line with each iteration, at the expense of losing performance cores with each iteration.

Noticed this, and it's kind of why I am glad I got an M1 Pro, because I think we're running into a situation where there isn't much left to squeeze out of these CPUs - regardless of benchmarks - unless your workloads are massively parallelized or highly dependent on GPU cores.

x64 or Apple Silicon. The only thing I do that taxes my PCs or Mac is video editing, and generally mostly in the GPU department. The CPU still has headroom.

But I'm not paying over $1K for a GPU (or $1,500+ to get comparable GPU power in an Apple build).

But decreasing Performance Cores in the "Pro" line is ... "very Apple."

I don't think "Pro" means the same to them as it does to us, though.

Lastly, the M1 Pro machines are already quite expensive, so this is probably done to draw a closer line for people to cross before they give up and just give them the extra money for a Max model.

All of Apple's Laptops are designed such that once you do a couple of upgrades, the price inches so close to the next model (which has many of the upgrades built into the base SKU) that you mind as well pay a couple hundred more for that one, instead. They probably make tons of money from people who do that.

When I bought my M1 MBP back when they released, it was over $2,018 ($1,899 + Tax) with 8->16GB RAM and 256GB->1TB SSD upgrades.

So, I waited and got the base M1 Pro when they went on sale for $1,499.