Apple introduces M2 Ultra

[Jac459]

It won't keep up with the 13900k.
Run Cinebench R23 multi-core test.

Hmm.. Cinebench is Video / 3D, which tends to have an easier time buffering so all cores are fully utilized. Real-time audio is a different proposition, and the more you lower the latency the less buffer available. So Cinebench isn't an ideal reference point, and shouldn't be used as a basis for DAW decisions.. IMO

Further, additional core types add more complexity to a hosts load balancing. Performance can (and does) vary greatly between hosts, and Cubase (what I use) is well known for various issues regarding core usage on Windows. I don't know how much they've "fixed" things, with the newest Intel CPU's, after its problems. But I'd still tend towards processors, with single core types, for Cubase under Windows. It's an example of a variable which mean certain choices might not be the best, regardless of what benchmarks like Cinebench say.

That'll tell you where the M2 Ultra stands vs a "workstation" type CPU.

It depends what you mean by "workstation". Many of Intel's Xeon's are silly money because they're not aimed at consumers. For DAW usage the M1 Ultra was competitive with the 12900k on launch. That's not just my opinion - DAWBench did limited cross-platform benchmarks which showed it beating the 12900k. M2 is only an incremental update to M1. It should gain a bit versus the latest Intel, but the real performance gains will come with the M3, which I'd expect (the Ultra) will beat the 13900k for DAW performance. Though, even if that's the case, the relative value for money is less - Apple would have to price the Ultra more like the Max to compete better there.

I use Cinebench as an example... because it's a quick/easy way to gauge both single-core and multi-core performance... and it's available for both platforms.
ie: If you look at multi-core scores for 12900ks (28k) vs 13900k (40k), you have a fairly accurate idea of DAW performance expectation. Certainly more relevant than synthetic benchmarks (Passmark/etc).

If you want to talk about less than ideal multi-core scenarios (especially at super low latency), that's where single-core high clock-speed will be a major factor.
You can gauge this by having a look at Cinebench R23 single-core results.

When I say "workstation" type CPU, I'm referring to a high-performance CPU... designed specifically for that purpose... knowing that it requires substantial cooling/space/etc.
I'm well-aware that Xeon CPUs are terrible for DAW purposes. That's due to low clock-speed.

The M3 may well best the 13900k... but I doubt AMD and Intel will be standing still.
AMD's next offering will likely leap-frog the 13900k (as did the 7950x vs the 12900k).
Rinse and Repeat.

The laws of thermodynamics are at play.
High clock-speed equates to high heat.
High IPC can mitigate that... to a point.
Right now, 7950x has slightly higher IPC vs 13900k.
13900k is slightly faster (overall) because it can achieve higher clock-speed... across more cores.
Both the 7950x and 13900k require substantial cooling.
Folks running them are choosing maximum performance over small-form-factor or convenience.
If Apple decides to push the performance curve, it won't be in a 10"x10"x6" form-factor.

Every week, I talk to Mac users who are essentially forced away (from Apple) because they can't get the performance/configuration they need to work (often composers for TV/Film).
Granted, this isn't a general-purpose computer user.
It's someone who needs ultimate high-performance... and maximum configuration flexibility.
I don't think Apple will get back into the highest-performance market... for the reason I said previously. They know small/sleek machines will sell far more to the general public.
DAW users are a tiny niche group. General-purpose computer users are a massively larger market.

Can you share were you find your benchmarks ? The ones I found have all a big "not verified" sign...

Normally for music producers it is also good to have silent computers no? What cooling technology are you using?

[revvy]

This thread be like:

-> Apple has a new processor
-> Their new AR headset is shite
-> Apple is a cult
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> You can't really make music on a Mac, they're too slow

[Jim Roseberry]

Every week, I talk to Mac users who are essentially forced away (from Apple) because they can't get the performance/configuration they need to work (often composers for TV/Film).

They must be doing something terribly wrong. I have yet to run out of headroom on the M1 Max with 32GB RAM. Huge template as well.

The scoring templates these folks are using could never be run with 32GB RAM.
RAM compression/cache used in Mini/Studio is not a substitute.

[machinesworking]

And you might notice that Jim said to run Cinebench R23 multi-core benchmark.
The results of which aren't Jim's opinion/bias.
The reason Cinebench exists is to gauge performance for Cinema 4D (3D modeling/animation).
If a CPU scores 28k, it's not the equal of a CPU that scores 40k.
Doesn't matter of your talking PC vs PC... or PC vs. Mac.

I did, I noticed Jim said run a 3d animation process in a forum dedicated to audio software, then ignored and continues to ignore that the same links showed in Geekbench multi core CPU suite tests the M1 was faster. So, which process score do you think is more likely to produce higher plug in counts across multiple tracks in a DAW? a specific number crunching test doing one thing unrelated to the kind of processes in a DAW? or a suite of tests?

I don't own the M1 Max or the other chip, and I'm not even that concerned about it, but you still refuse to address my original point, in some tests the M1 is faster, and I find that annoying since you obviously know what you're doing with computers. You sell them, and refuse to even talk about the strengths of the M1 without attempting to diminish them by referring to them as mobile chips. Which is a straw man of epic proportions considering the origins of the Centrino, i7 etc. etc.

Look, Intel and AMD have 64-128 core chips, we're all aware their market includes servers that Apple has seemingly no interest in providing for. Apple runs a different OS with different ways to boost performance that can be ignored or used by various developers, so why is this discussion even important? IMO it's for the same reason that people a few years ago would talk about how if you ran Windows on an Intel Mac you would get 100 instances of Reaktor instead of 98, as if in the real world a 2% increase in performance is worth anything at all.

This is KVR though, so any thread on something new and interesting from Apple is going to get people trashing it, it's a given.

[chk071]

This is KVR though, so any thread on something new and interesting from Apple is going to get people trashing it, it's a given.

If the Apple folks even deny the most factual and objective stuff like benchmarks, then there's not much to argue about indeed. I don't argue with the Jehova witnesses either.

I just wonder why the Apple clientele thinks that Apple is such a magic manufacturer that it can produce CPU's which are very energy efficient, and still beat any maxed out desktop CPU out there. That's a contradiction. You either have something energy efficient, or something performant. Or, you use your magic wand, and produce something which doesn't exist in the real world.

It's always the same as well. When I told some people here that my mid tier i5-11600K is about as performant (maybe a bit less) as a M1, I was called a liar as well. The benchmarks speak for themselves though. And, that's the stuff that matters in the real world.

Apple's M chips are amazing mobile CPU's. But, that's it. They're still mobile CPU's, and not something which can (or should) compete with the high end of desktop CPU's.

[Etienne1973]

I don't argue with the Jehova witnesses either.

As a happy Mac user for 30+ years: I find it amusing that most of us allegedly just follow a cult. But benchmarks are benchmarks, agreed. Apple Mac Pro should deliver. Period. Perhaps M3 will but AMD/Intel won't stand still.

[PAK]

If the Apple folks even deny the most factual and objective stuff like benchmarks, then there's not much to argue about indeed.

That's a bit vague. Who was denying what? Are you actually arguing that Cinebench is anything more than entirely useless for gauging the performance of DAW software?

You either have something energy efficient, or something performant. Or, you use your magic wand, and produce something which doesn't exist in the real world.

Ok. I already linked to benchmarks (from DAWBench) where an M1 Ultra outperforms the 13900k in Kontakt at a 128 buffer. Now please explain, in as much technical detail as you can, why it does this if what you say is always correct?

Or, just maybe, you've formulated a slightly over-simplistic argument? Surely not..

[chk071]

I don't argue with the Jehova witnesses either.

As a happy Mac user for 30+ years: I find it amusing that most of us allegedly just follow a cult.

Not every Mac user, of course. I wouldn't call my own mother a Jeohva witness. There definitely seem to be some very hardcore Mac users though. Almost worse than those Linux fanatics.

[chk071]

But benchmarks are benchmarks, agreed. Apple Mac Pro should deliver. Period. Perhaps M3 will but AMD/Intel won't stand still.

Frankly, I think the Mac folks can be happy that they already achieve the best mobile processors. If they achieve desktop CPU like performance with those, then Intel and AMD surely will have to ask themselves WTF they're doing all day long.

But, that's fairytale stuff. They simply won't.

[chk071]

You either have something energy efficient, or something performant. Or, you use your magic wand, and produce something which doesn't exist in the real world.

Ok. I already linked to benchmarks (from DAWBench) where an M1 Ultra outperforms the 13900k in Kontakt at a 128 buffer.

Well, that's asweome, but, I rather rely on benchmarks which cover many use cases, not just virtual instruments, which may or may not be optimized for the one or the other CPU extensions.

[PAK]

Well, that's asweome, but, I rather rely on benchmarks which cover many use cases, not just virtual instruments, which may or may not be optimized for the one or the other CPU extensions.

Ok, valid choice. But at least acknowledge it contradicts what you said somewhat?

The reality is DAW software, almost more than anything else - thanks to the host + plugins structure, has a bunch of variables in play, and Apple Silicon can be a bit of a mixed bag itself. Whilst Cinebench benchmarks will tell you to steer clear, for those use cases, the performance delta isn't of the same magnitude for many audio purposes.

What your scenario, regarding performance, leaves out is that the overall package matters. EG Apple stuck all of its memory on its CPU. Which, in one sense, is annoying. But, in another sense, now allows them to have way more bandwidth (up to 800GB/s for M1 & M2 Ultra) than present PC system memory allows. Whilst the primary benefit of RISC has traditionally been less energy use, in theory, this also means they're left with more headroom if they actually want to push the megahurtz..

Chips are now being designed with specific functions in mind too. The most obvious example being video codecs (which caused some to over-hype the performance of M1 etc) ..But imagine if they start targeting stuff, like that, at audio software

[Jim Roseberry]

I have a M1 max
I have a 7950x
I have a 13900k

If you're talking raw performance, the 13900k is the fastest CPU.
Doesn't matter if you're running DAW applications, Cinema 4D, Video editing/rendering, etc.

3D Modeling/Animation makes multi-track audio look lightweight.
I mention Cinebench because most folks aren't going to take the time to do a full-scale DAW test.
Yes, Cinema 4D (Cinebench) is a better DAW performance benchmark than purely synthetic benchmarks.
ie: The 7950x bests the 13900k in 7-Zip compression. Does that mean the 7950x is faster than the 13900k?
Nope. The 13900k is slightly faster for most scenarios (including DAW).

I never said someone couldn't make music on a Mac.
If I had the likes of a M1 mini when I was a kid, I would have thought it amazing.
Someone asked how the M2 would compare to the 13900k.
I mentioned an easy way to gauge heavy multi-core performance.
To Apple's credit, they don't tout or market the M1 or M2 as a "workstation" or "power-user" type CPU/machine.
It doesn't take a genius to figure out their marketing/direction.
If Apple were wanting to push the performance envelope, it wouldn't be in a small form-factor.
They'd be cranking clock-speed to highest possible (like AMD, Intel)... requiring large cooling/cases.

Apple managed to squeeze a lot out of the M1 mini.
For someone who's a professional composer... running a large scoring template, it's not practical and doesn't work.
No, these folks aren't doing anything wrong.
They're not the user writing songs... making moderate use of Virtual-Instruments.
Many of these scoring templates are consuming most of 128GB of physical RAM.
Ram compression/cache (as used in Mini) would kill performance. I've tested this.
Apple released the Studio... which you can get with 64GB/128GB physical RAM.

Apple has made a more broad-based financial decision... choosing to cater to a much larger target market.
Say the M1 Mini has outsold the Mac Pros 50:1 (I'd bet that's being conservative)
Say the M1 Mini has a street price of $800
Say the Mac Pro has a street price of $5000
Let's say the profit margin is 30%
M1 Mini: $800 x 50 (pieces) = $40k gross... $12k profit
Mac Pro: $5000 x 1 = $5000 gross... $1.5k profit

For better or worse... like it or not... this is why Apple isn't catering to power-users.
It's a sound business decision.

[vurt]

This thread be like:

-> Apple has a new processor
-> Their new AR headset is shite
-> Apple is a cult
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> Their new CPU won't be as fast as Intel/AMD
-> You can't really make music on a Mac, they're too slow

took the heat of the imask though!

[chk071]

Well, that's asweome, but, I rather rely on benchmarks which cover many use cases, not just virtual instruments, which may or may not be optimized for the one or the other CPU extensions.

Ok, valid choice. But at least acknowledge it contradicts what you said somewhat?

It doesn't as it's a single case scenario, with a single plugin. I have no idea whether or not Kontakt is optimized to work better with Apple's ARM CPU's. It doesn't say anything about the actul performance of the CPU compared to others.

Read Jim Roseberry's post above, it explains why it's nonsense to just take a single plugin and test it with two CPU's. Well, maybe not nonsense if you exclusively want to use that plugin, and it's all you will ever use. But, nonsense in terms of determining the performance of a CPU in comparison to others.

[Etienne1973]

I don't argue with the Jehova witnesses either.

As a happy Mac user for 30+ years: I find it amusing that most of us allegedly just follow a cult.

Not every Mac user, of course. I wouldn't call my own mother a Jeohva witness. There definitely seem to be some very hardcore Mac users though. Almost worse than those Linux fanatics.

Lately some Mac fans were raining on the parade in the Intel 13900k thread here at KVR Audio. I think that was wrong. But WHAT are you doing here in an Apple Mac thread?