New "Mac Studio" from Apple

[Eric Lambert]

Mine arrives tomorrow. I'll probably have a lot to say over the next few weeks.

[kings79]

Originally Posted by Eric Lambert

Mine arrives tomorrow. I'll probably have a lot to say over the next few weeks.

Keen for your thoughts Eric

[audiolex1]

Originally Posted by kings79

If some orange headed merican dude yabbering about geekbench scores is enough to quash real world usage experience, then ok. I wave the white flag.

Also man. he did go on to say the Ultra was overkill for most DAW users. Also man. He didn't even have a Max to do any kind of 'test' like you said. Zero tests... Not that I would take what Barry Johns Studio Talk as gospel anyway.

Did you watch the video in full?

ALL I"M SAYING IS: The Ultra DOES NOT automatically give you double the speed in real world usage. Maybe rendering video cause there are chips specifaclly for that on the SOC but you just don't get double the speed for what we do. I NEVER said that Audio was single core ONLY. I just said that is ground zero and where you're basing the whole chip series from.

Anyway. if you wanna spend double your hard earned on the Ultra. Go right ahead.

All I am saying is he said the single score didn't mean much. AS they were near identical, the multicore score is where things changed.

And I think that is the point. the x86 architecture it seems to mean more then the ARM where multi core means more.

I don't think the Max would have made that much of a difference in the test as he covered most of the bases.
The Max just has more GPU cores, which isn't going to do much for the test. The Ultra already had that and the M1 Pro is basically the same as the MAX in cores.

So that is overly redundant. As for "doubling", I'd say being able to go from 50 tracks to 170 tracks is a little more than doubling.
You are equating speed/performance vs count or the ability to do more.

If you track at 32 buffers all the time, I'd say that would make a difference. And he was overloading on the plugins that in a real wold situation would be overkill..

And just to restate. Max ONLY adds 8 more CPU cores. So the test would have been a waste of time.
And he is probably right that the Ultra is a bit much for most home studio users, but his point was if you needed more power, it's there in the ultra.

I've notice this with the Intel base users. They hate that the M1 is beating the Intel and the old CISC and they aren't catching up either.

This is only the beginning. Once everything goes Native and unlocks the full ARM architecture its going to leave Intel/AMD behind.
And Apple hasn't even unlocked the 32/64/128 cores yet.

[kings79]

Originally Posted by audiolex1

All I am saying is he said the single score didn't mean much. AS they were near identical, the multicore score is where things changed.

AND

ALL I"M SAYING IS: The Ultra DOES NOT automatically give you double the speed in real world usage. I NEVER said that Audio was single core ONLY. I just said that is ground zero and where you're basing the whole chip series from.

[Eric Lambert]

I think this is still a tributary from a sentence I'd written upstream, which did not claim that the Ultra provides double the speed, only double the power. Let's assume I'd said double the "potential" instead, because everyone seems to agree that DAW usage is enigmatic and complex and the Ultra isn't twice as fast with ProTools.

Ignoring specific metrics and percentages, it does, however, bring more to the table for ProTools today, and it will bring much more tomorrow (for me, literally tomorrow). And it will be overboard for some right now, and not for others.

I did some math and calculated what my current MacPro has cost me per day and it comes to... one dollar! $1/day. After almost a decade, that machine -- overpowered when new -- has lasted so long that it's prorated to $1/day. If I can get anywhere near that with the new Studio I'll be absolutely thrilled, and I think the Ultra will last much longer. I could be wrong, but I took the bet anyway.

[kings79]

Originally Posted by Eric Lambert

I think this is still a tributary from a sentence I'd written upstream, which did not claim that the Ultra provides double the speed, only double the power. Let's assume I'd said double the "potential" instead, because everyone seems to agree that DAW usage is enigmatic and complex and the Ultra isn't twice as fast with ProTools.

Agree 100%

Potential. Geekbench blah blah blah.

the Ultra isn't twice as fast with ProTools

But this is a PT thread and we're all using and talking about Pro Tools. Arn't we? Or am I in the wrong forum again?...

[Michael Carnes]

I've collected some real world data for you, using my M1-Max MacBook Pro. The goal was a decent mock-up of an orchestral piece I finished recently. The piece is a short 3'45" and in the real world would use about 80 players or so--depending on the size of the string section. I'm going to try to avoid mentioning brands of things--suffice to say that I'm using VIs that are quite good and quite popular. I'll be respectful of our Avid hosts here and refrain from mentioning the DAW I used. Suffice to say, it's a DAW that can run both Apple Native and under Rosetta. My goal was a 7.1.4 mix. I'll turn it to Atmos later on, so the intermediate goal was simply the discrete mix in 7.1.4 at 48K.

I ended up with about 140 tracks. Most are stereo VI tracks. There are a few summing busses, an immersive reverb that runs both native and Rosetta, and some VCA controls. While my goals are not to make a film or TV score, this is probably in the neighborhood of what you might need for a small film or medium TV score.

The first number I have for you is the amount of memory taken up by my DAW (which includes the preloaded first bit of each instrument). As you know the preloaded part allows notes to begin right away as the rest of the sound is streamed into memory. Running Native, the project took up a little over 47GB. Under Rosetta, it took a bit over 51GB. The difference is probably the Rosetta code itself, plus whatever difference the Intel code might make. In either case, you can see that I'm using up most of the 64GB available in the M1 Max. The remainder will probably go to the Dolby Renderer when I hook that in. This amount is a little close for comfort. I'm only using a single mic position for each VI. If you like a template that gives you the option of playing with multiple positions, then you might go over this memory limit unless you freeze tracks as you work. For those of you who like having a large template as a starting point (even if you hone it down), you'll probably find this limit a little uncomfortable.

The second number is CPU load. The Max has 10 cores--2 'efficiency' cores and 8 power cores. Since all of them get used, I'll treat them as equivalent (even if that's not quite true). Running this 225 second piece in realtime took 404 CPU seconds in Native and 679 CPU seconds in Rosetta. One way of looking at this is that it required about 2 full cores in Native and required about 3 full cores under Rosetta. Of course the real processing was spread over all 10. The conclusion is that CPU power isn't really much of a limitation. In either case, I was using much less than half the power available to me. I should point out that I don't use a lot of plugins many of you are likely to use. For me, it's just verb and a touch of EQ or dynamics when needed. If you have other goals, you're likely to use a lot more of that CPU power.

I should point out that my sample libraries are all out on T-Bolt M.2 RAIDS (3 of them). That gives me a huge amount of bandwidth for getting samples into a job. This streaming barely made a dent in the system, so I didn't spend much time looking into it.

It's pretty clear to me that the limiting constraint of the M1 Max is memory. I'll run out of integrated memory (RAM) long before I run out of CPU. Of course that will change. The available power means--as it always has--that DSP can improve. Perhaps we'll see more modeled instruments too. I'm not sure about the graphics cores. They'll certainly allow UIs to look nicer and be quicker. I also expect that a lot of offline processes (noise removal, etc) will begin to take advantage of them. Machine learning developers love doing a lot of stuff on graphics cards. With the Mac GPUs now being always available, I think we'll see them becoming part of the audio developer's toolkit.

So there really is a place for an Ultra in some music/post applications. The Max does really well, but I can see where the limits are. Hope this is helpful.

[K Roche]

Very Interesting Micheal and helpful,, thank you for posting

And in a general real world use sense I am curious ::: when twice the money for computer has ever,,, equated to twice the speed in Pro Tools ? Or why that would even matter

[Michael Carnes]

Originally Posted by K Roche

Very Interesting Micheal and helpful,, thank you for posting

And in a general real world use sense I am curious ::: when twice the money for computer has ever,,, equated to twice the speed in Pro Tools ? Or why that would even matter

That's a good question. I might draw the analogy that twice the money for a fancier car might not get twice the speed, carrying capacity, lowered carbon footprint, etc. But--if you're careful in choosing--it will get you a better car.

Some apps are written for a single core because that's all they need. Or they don't easily factor out into multiple threads. In that case, single core performance is the only metric. High-end audio apps like Pro Tools and every other one I can think of are all written to distribute the load across as many cores as are available. For simpler workflows in realtime, you're not going to notice any improvement as long as the system is fast enough for that workflow. You are likely to notice the difference for offline rendering though. There's a great old term from the early days of realtime programming: "Fast enough is fast enough".

I think the difference that a bigger machine like the Ultra offers is the ability to do bigger projects: more tracks, more VIs, beefier plugins etc. More means bigger more than it means faster. I do think that as these systems reach critical mass in the user base you will seen new sorts of programs arise. That's because there's a guaranteed basic pool of processing units, whether CPU, GPU or Neural. I'm not likely to explore that very much as a developer. I'm enjoying a retirement that lets me be a musician again. But I can't wait to see what those young pups come up with.

[K Roche]

Originally Posted by Michael Carnes

That's a good question. I might draw the analogy that twice the money for a fancier car might not get twice the speed, carrying capacity, lowered carbon footprint, etc. But--if you're careful in choosing--it will get you a better car.

Some apps are written for a single core because that's all they need. Or they don't easily factor out into multiple threads. In that case, single core performance is the only metric. High-end audio apps like Pro Tools and every other one I can think of are all written to distribute the load across as many cores as are available. For simpler workflows in realtime, you're not going to notice any improvement as long as the system is fast enough for that workflow. You are likely to notice the difference for offline rendering though. There's a great old term from the early days of realtime programming: "Fast enough is fast enough".

I think the difference that a bigger machine like the Ultra offers is the ability to do bigger projects: more tracks, more VIs, beefier plugins etc. More means bigger more than it means faster. I do think that as these systems reach critical mass in the user base you will seen new sorts of programs arise. That's because there's a guaranteed basic pool of processing units, whether CPU, GPU or Neural. I'm not likely to explore that very much as a developer. I'm enjoying a retirement that lets me be a musician again. But I can't wait to see what those young pups come up with.

Not to mention that many PT user (myself included )who use to only produce audio now also work with video