Just got an m4 Pro. I know I’m still installing and waiting on downloads etc, but I just opened logic and did a quick test, and ozone on the master bus was causing maxed out single core at 128 buffer with some soothes in series into it.

I upgraded from m1 mainly for single core reasons. I use a lot of bus processing and hate high buffer. I was under the impression the m4’s single core was almost twice as fast and would result in a huge step up for bus processing / buffer (since bus processing is all done on a single core)

I know I know I should wait a few days til the computer stops installing and indexing etc. but I am having trouble finding any real world results from people that push single core pretty hard and how much of a real world improvement or not the m4 is. Anyone?

A bit in here
https://www.youtube.com/watch?v=LaqSMSRzLe4

I have an M4 Mini Pro. I can assure you that It's stupid fast, and the single core performance is what you hope it is.

BUT...

It's a new machine with a new OS. A few plugins don't play nice, and slow the whole thing down. Anything that's not Apple Native, just get rid of it. Also, some plugins can slow down all plugins GUIs (usually turning off graphics hardware acceleration helps). And there's some that the AU version is fine, but the VST isn't. And vice versa.

For the others, update to the latest.

Don't worry, it's usually just a handful that cause issues. Which handful is the question, and you can only sort it out by testing.

Now, I'm running Studio One, not Logic. But I'm able to load it up to the teeth at a 32 sample buffer. And Logic should behave the best on a new system, given it's Apple's product. Also, if your interface has a driver that's not happy with M4 or Sequoia, there's not much you can do.

What you might want to try is using only Logic plugins as a test, and see how low you can get the buffer.

For my favorite benchmark my M4 is 69% faster than my M1.

Go into the audio preferences in Logic and make sure the Processing Threads is set to the highest number, not automatic. By default Logic seemed to only be using the P cores in automatic mode, and once I switched it to the highest number (10 in my case on an M4 mini), the performance of Logic as a whole increased greatly. I know you’re talking about single core performance, but doing this this might take some of the load off of the P cores in your project and help it to handle your master bus processing better.

Yup, I always do this. However, I find it's not good to use the E cores... For me, it slows things down system wide and causes more issues than solves.

However, after a day of things loading, Logic is a dream. Sessions that were completely crapping out on the M1 Pro are running like a dream with 30-40% CPU, and no overages. 👌

Amazing! Love to hear it.

I just bought a 16” MBP M4 Max and recently before that an Intel 13900K system.

Both machines have a single core performance of about 2.5x that of my 10 year old 4th series i7.

So as you can see single core performance hasn’t come very far in the last 10 years.

And now they’re at 3nm dyes and the transistors are a few atoms apart and cycling at near melting speeds …. I don’t see a very exiting road map ahead for silicon technology based CPU’s and sc performance.

Sure the width with 24 cores etc is amazing - hundreds and hundred of tracks … but real time VI’s at low latency … it’s never really got there for me.

The RAM and drive speeds are great and libraries load up like switching TV channels but single core performance - it’s got nowhere to go with a single core of silicon already at 3nm.

Maybe one day they’ll figure how to link cores together without bridging latency.

Hmmmm. You sure about that. According to geekbench, 4th series i7 has single core score of 1200. M4 has 3900-4000. That’s over 3x. Also my m4 I’m noticing a huge single core jump even over my m1 now that the dust has settled.

Geekbench doesn't fall under "real world".

Definitely doesn’t translate into real time audio.

So you guys are saying, a brand new M4 Pro chip will not be 3.5x faster in real world than a 10 year old i7? Based on my observations, and comparing to my M1 which already was a big jump, I disagree. I would like to see an i7 handle 1/3rd of the bus processing I'm doing on the m4 right now. I bet real world is in fact not far off from the geek bench numbers.

I missed that you said 2.5x faster single core here. Looking more like just over 3x. But, in my opinion that is a pretty huge jump even over a decade, considering we often get used to more incremental changes and single core is the hardest to deal with.

There’s literally a thread showing the M4 performance. It doesn’t perform to the same level as Geekbench would have you believe. It’s no slouch. But it’s no king like Geekbench “shows”. Then again, you can’t get the 9950x or 285 in a laptop without it catching fire!

I’m sure it’s better than the i7.

That's why I asked at the top real world opinions. My first proper day today is showing that the m4 pro is in fact an insane jump. So I'm not sure what threads you are referring to, can you point me to them?

Nobody is saying it isn't a big jump. All I said was that Geekbench didn't seem to translate into DAW performance.

Not sure, I mean this is comparing to Intels desktop chip. And it holds up remarkably well and is in a laptop and silent. So I'm not sure this proves anything related to geek bench and how it translates to real world usage.

I guess I can’t help you then. Good luck.

One man’s “pretty huge” is another man’s “surprisingly small”

Compare sc performance increase to

RAM Bandwidth
RAM size!!
HD to SSD’s to NVMe speed!!
Core count 4 > 8 > 12 > 16 > 24 > ??
GPU speed, RAM, bandwidth resolution.
Monitor res. 768 > HD > QHD > 4K > 8K
Monitor size 17” > 24” > 32” > 75” !!!!
Broadband 7kb > 100Kb > 1mbs > 1Gbs

Then we come to the single core.

And the the context of the above I don’t think it’s made very much progress. Nobody’s fault of course - it’s just physical limitations.

The height has gone up about 2.5x
The width (multiple cores) has gone up about 10x plus.

Sure we can run hundreds of tracks and thousands of plug-ins.

But stick on an Orange tree samples electric guitar (superb by the way) in Kontakt 8 put your buffer at 32 at 96KHz and press the sustain pedal for some fluid chord arppegios and …. oops - overload.

Try 64 buffers at 48KHZ …. Might be a bit better still falls over depending on other processing.

128 / 48KHz and I’m good.

So whilst I do greatly admire the tech behind my new MBP M4 Max and 13900K PC workstations - I’m personally waiting for some paradigm change in the CPU world away from the concept of a single core of silicon.

Quantum computing maybe?

It depends on the coding. There can be instructions where the new CPUs are many times faster than the 10 year old i7.

With other instructions, the delta may not be as large.

“Real world” means different things to different people.

If all plugins were coded the same way then it would be easier to compare things - but they’re not.

Wasn’t it @ TNM that compared IK tape sims and got an insane delta? But other plugins weren’t the same delta.

For example: amazing coding on x64 but a sloppy port to apple silicon? You probably won’t see a major improvement.

Conversely, if someone spent the time utilising the power properly on the new chips, they could get some amazing results.

Geekbench is helpful at showing the improvements generationally for microbenchmarks. Look at the results and they show the performance for doing different things, most people look at the big final number but it’s worth looking at the individual scores of the microbenchmark components and track the difference that way.

That would all be a lot more impressive if you were giving the specific model number of that "i7". You don't even say whether it is a desktop or laptop CPU.

FWIW for my benchmark the M4 is 2.4x faster than a i7-4900MQ.

That’s coding.

Sure we always need to use our tools more “efficiently”

I’m talking about hardware.

They’re at 3nm now - transistors are literally atoms apart.

A core is just that - a single silicon CPU.

That tech has reached the end of the road as far as I can see.

I mean, don’t get me wrong - I can make music on these modern systems but the days of “buffer juggling” are never going to be over until we get some new tech.

My last PC and Mac lasted me 9 years - until they come of with something new I’ll be using my new systems for 10 years plus!

A good thing from a wallet point of view

From the sheer standpoint of how many transistors you can pack in a given amount of space, I agree we are approaching the end of the road without some major change in tech.

But that’s only one aspect of it.


How some instructions are executed can always be further optimised. Special processing can be added to chips (as Apple has done with dedicated ProRes engines) that massively accelerate workflows compared to the past.

So even if the single core can’t get denser or run at a higher clock speed, if it invokes instructions that are 10x faster vs their inefficient old selves, it will go 10x faster as far as you’re concerned

We are far from being “done” yet with that part of CPU design.

We are starting to see NPU being added to CPUs. Look at an M4 and it has dedicated silicon for offloading various things so that IF someone takes advantage of the hardware, you can massively ease the load on the main cores.

Apparently some stuff like Chromaglow in logic use the NPU, which is a nice first step.

Perhaps you’re right.

I’m waiting but I’m not going to hold my breath on this.

I still think something new will come along and make a single core of silicon look quite basic in concept.

because if you look at the geekbench scores for these 3 CPU's the M4 should be the best ..... and it isn't in Real World DAW related tasks.

He also mentions the M4 gets very hot when pushed to these limits and the fans ramp unlike the larger form factor intel and AMD which stay very quiet at all loads do to them being able to use large cooling AIO's.


yes, they're amazing lapatops if that's your thing but this is about geekbench not being a reliable bench of how a CPU performs in a real world DAW environment.


M

I did my own benchmark test where I tested Single core performance by stacking Logic Pro's ChromaGlow plugin.
I compared my MacBook Air M1 with the new entry level MacBook Pro M4.

The M4 handled 35 percent more load in my test.

In comparison, the M4 has 37.5 percent higher processor speed than the M1.
But according to Geekbench, the M4 should be over 50 percent better!

For me, it was disappointing that the M4 did not deliver better results.

Ya, all really good valid points here. I'm sure Geekbench isn't perfectly accurate across chips, applications, etc. However, within apples own ecosystem, it's a pretty accurate representation of comparable processing.

I've been playing last 2 days with my m4 pro, and it destroys my m1 at low buffer single core. I found with the M1, right out of the gates, I was having to run at 512 buffer pretty early on in sessions due to cracks and pops. Today, I was running a bunch of stuff, and tracking electric guitar, software monitoring through Guitar rig with some heavier presets, at 64 buffer, while duplicating tracks during playback and record enabling them and continuing on like this, without a single hiccup. Also, was able to run sessions that were crapping out at 1024 Buffer on my M1, at 128 on my m4 no problem, with lots of room to spare.

So for me, the real world experience of M4 vs M1 is a very large jump in single core experience. I'm very excited to be able to run at lower buffer later into projects as they get heavier on CPU. This alone is a game changer for me

Oh, yeah? Is there an Apple machine around that M4 Pro and exactly which one is it?

No machine, just the m4 chip. Not sure why it’s not working!?! 😂

Yes, MacBook Pro 16”