M3 - real Live Frequenzy ? is it not 4.05Gh ? whats going on ?

[streetfunk]

Hi there,

in several Max-Tech Vids vs. Benchmarking the new M3s,
it became directly visible that these M3 macs do not necessarily run at 4.05Ghz speed.

I´d like to know more about it.
Whats going on ? is this normal ?
When would these M3 macs go to full clock speed, and when not ?

here is a YT-video for example, check from 9:16 on.

personally i´m concerned about my music making.
My main app runs only ST.
The task is realtime-playing.
The CPU load is for example also determed by how many notes i play.
So, the whole CPU load i do generate is something very fluid !
I run my M2pro vs. my ST task into its knees.
I *really* need SC speed !

do i have to be concerned ?
Is this the same with the M2pros ?
How can i look the CPU speed of my cores in realtime ? What app do i need to have ?

 

[theorist9]

The Apple Silicon (AS) Macs have dynamic frequency scaling. [Intel has the same thing—they call it "Turbo Boost".] When you are significantly loading only one core—which happens when you are running a single process with nothing else significant going on—the M3 can run at its max of 4.05 GHz. But as more cores come online (e.g., you've got Spotlight indexing going on in the background), the clock speed decreases. According to https://www.notebookcheck.net/Apple...ormance-and-improved-efficiency.766789.0.html , the M3's frequency when all cores are loaded decreases to ≈3.6 GHz.

Are you saying that your M2 Pro isn't able to keep up with your single-threaded music app? If so, an M3 probably wouldn't help, because its single-core Geekbench 6 performance is only ≈14% better.

Does your app run natively on AS? If not, that can hurt performance, and you might want to look into getting one that is AS-native.

Or if that's not the issue, I'm wondering if you can find either a more efficiently-coded music app—or one that can make use of multiple cores. I'd suggest posting on the digitial music forums to ask for suggestions.

 

[mecloud]

personally i´m concerned about my music making.
My main app runs only ST.

I’m curious what app this is. Most of the music-production apps I’m familiar with are highly multi-threaded.

 

[streetfunk]

The Apple Silicon (AS) Macs have dynamic frequency scaling. [Intel has the same thing—they call it "Turbo Boost".] When you are significantly loading only one core—which happens when you are running a single process with nothing else significant going on—the M3 can run at its max of 4.05 GHz. But as more cores come online (e.g., you've got Spotlight indexing going on in the background), the clock speed decreases. According to https://www.notebookcheck.net/Apple...ormance-and-improved-efficiency.766789.0.html , the M3's frequency when all cores are loaded decreases to ≈3.6 GHz.

Ahh, i see ! This helps to understand ! Much Thanks

Are you saying that your M2 Pro isn't able to keep up with your single-threaded music app? If so, an M3 probably wouldn't help, because its single-core Geekbench 6 performance is only ≈14% better.

Its a plugin Host ! the load depends on me vs. what i "patch" .
Yes, it´s "initself" just single threaded.
No, 14% more does in fact help. It helped from M1 to M2, and will also from M2 to M3.
If i get those ~14%, at least . Which was my concern now.
Yes, i load one core to full extent. Easily in fact.

Does your app run natively on AS?

Yes it is. It´s in fact both. I use it both ways. ARM native and under rosetta2.
Depends if i have to use plugins which are still not AS ready, or not.

I'm wondering if you can find either a more efficiently-coded music app—or one that can make use of multiple cores.

More efficiently coded ? Thats not available. It´s in fact the leading app on that ground.
Nothing available that could anything compete with it for my uses.
And so far, the Devs keep it ST by intention. It´s by will.
see =>

I’m curious what app this is. Most of the music-production apps I’m familiar with are highly multi-threaded.

This is "GigPerformer". It´s a plugin-host, specifically tailored towards live on-stage uses.
It´s absolutely the leading one -for those in the knows- and it is slowly slowly turning to become the defacto standard. At least for the live onStage people.

Outside of the live on stage scene, is it not very well known, respectivly just overlooked.
But it is the goto thing => when it comes to Audio-plugin hosting for any sort of "live-play uses".
There, it outperforms anything else vs. most of the specific use cases ! It is running more stable than anything else and has less latency too. On top of its wonderful -under the hood- feature set.

Yes, the DAWs are optimised for multi-threading.
But when it comes to realtime-play usages, are the DAWs just not on par.
It´s a different type of work.

Cheers, and thanks for feedback !

 

[theorist9]

Depends if i have to use plugins which are still not AS ready, or not.

More efficiently coded ? Thats not available. It´s in fact the leading app on that ground.
Nothing available that could anything compete with it for my uses.
And so far, the Devs keep it ST by intention. It´s by will.
see =>

This is "GigPerformer". It´s a plugin-host, specifically tailored towards live on-stage uses.
It´s absolutely the leading one -for those in the knows- and it is slowly slowly turning to become the defacto standard. At least for the live onStage people.

I was curious how it could be the leading app when it brings even an Apple M2, which has one of the fastest SC speeds on the market, "into its knees". That makes it sound like it's not very usable.

So I checked the GigPerformer site, and found this thread:

Apple Silicon Chip Comparison

I was doing a bit of research trying to decide which new Mac to buy and ran across this video. Thought some of you might be interested. Very illuminating to me. I wonder what the CPU usage of efficiency vs performance would be for GP? It was very different for the DAWs he tested in the video.

community.gigperformer.com

The first responder got an M2 Max with 96GB memory and says the machine is more powerful than he needs. So these possibilities come to mind (just throwing these out there):

1) You don't have enough memory (no idea if this is the case, but it's worth looking into). You can check this by seeing if the SWAP increases when you are using the program, or simply by monitoring the memory pressure (you want it to stay green) in Activity Monitor.

2) You are using less conservative plug-in settings (sampling rate, buffer settings?), or otherwise using the machine in a way that is more intensive.

3) Some of the other plug-ins you are using at the same time aren't AS-native.

4) You said it's a plug-in, which means it plugs into another app. What app does it plug into, and might it work better if plugged into a different app?

In addition, further down, it's mentioned if you have more than one instrument you can create multiple instances, which distributes the load across multiple cores. However, they say it's tricky to manage and might not be advisable for a live performance.

 

[leman]

To put this in perspective - M3 family running on all cores should maintain pretty much the same frequency as M2 Max/Ultra at peak single core. That’s a very decent generational improvement.

 

[streetfunk]

4) You said it's a plug-in, which means it plugs into another app. What app does it plug into, and might it work better if plugged into a different app?

ahh, not sure if i misspelled something ?
Gigperformer (GP) is a plugin Host. i run plugins within it.
GP can´t run as a plugin on itself.
There are some other hosts which can do that. GP can not.

2) You are using less conservative plug-in settings (sampling rate, buffer settings?), or otherwise using the machine in a way that is more intensive.

Yes, i do some other things.
Most of the guys there are live-onStage Band people.
Some few are Guitarists for home purposes.
The band people have to serve a band context, they use completly different setups.

What i am doing could you consider to be realtime-play sounddesign.
This involves way higher amounts of plugins beeing in use at the same time.
I work based on using FX. (plugin effects)

Some FX have to be oversampled to make them sound good.
Some FX have to be loaded into another host first, a host that allows to set single plugins to oversampling, a host which can be run as a plugin within GP. So it can happen that i have to nest hosts, and this several times within one GP patch. To make the things possible i´m doing. This is realtime-play sounddesign.
Just a very few years back would even hollywood people have told you, "thats not possible" ....not in realtime....what we can do right now.

Its that specific piece of software which makes some major aspects possible.
But GP is running Single threaded only. Thats the caveat.
so, no, now workarounds possible, based on using other pieces of software.
Its the forefront.

In addition, further down, it's mentioned if you have more than one instrument you can create multiple instances, which distributes the load across multiple cores.

yup, that´s doable ! I do this !
but things are then becoming unfunny to handle.
And more prone to failures. And it adds more latency.
It quadruples the workload for me

But thats a cool feature to run several instances in parallel.
I run in fact allways two instances at a time.

1) You don't have enough memory (no idea if this is the case, but it's worth looking into). You can check this by seeing if the SWAP increases when you are using the program, or simply by monitoring the memory pressure (you want it to stay green) in Activity Monitor.

i run 16GB of RAM. thats plenty fine for what i´m doing !
even 8GB would be enough for this task.
then just close the browser and other apps.
So, RAM is here not a big thematic. At least nor for me

 

[streetfunk]

To put this in perspective - M3 family running on all cores should maintain pretty much the same frequency as M2 Max/Ultra at peak single core. That’s a very decent generational improvement.

i agree !
i have an eye towards the 3max

but as sayed, my main music app runs only on one thread. Thats kept so by purpose.
I think, running without any compromise for biggest system stability possible is here the main driving factor, for the devs. The piece of software is strictly made for live onStage gigging musicans.
System stability is the #1 "thing" here. It´s absolutely understandable.

 

[cupcakes2000]

Unrelated, but what do you do exactly? Sounds cool!

 

[JustAnExpat]

I decided to take a peak at Gigperformer's website at https://gigperformer.com/ . I never used the software, and I'm not going to download it (I'm not sure how to use it actually!). However, there is a line that I noticed:

"Gig Performer 4® runs on both OSX (Universal for Intel and Apple Silicon) and MS Windows so you can enjoy its benefits on whichever platform you are most familiar."

Is this a native Apple Silicon application, or is it running in Rosetta mode? Can you open up Activity Monitor (Applications/Utilities/Activity Monitor.app) and see if it's an Intel application running under Rosetta or a native Apple Silicon application?

If it's a native Intel application running under Rosetta, it could be the overhead caused by Rosetta.

 

[theorist9]

I decided to take a peak at Gigperformer's website at https://gigperformer.com/ . I never used the software, and I'm not going to download it (I'm not sure how to use it actually!). However, there is a line that I noticed:

"Gig Performer 4® runs on both OSX (Universal for Intel and Apple Silicon) and MS Windows so you can enjoy its benefits on whichever platform you are most familiar."

Is this a native Apple Silicon application, or is it running in Rosetta mode? Can you open up Activity Monitor (Applications/Utilities/Activity Monitor.app) and see if it's an Intel application running under Rosetta or a native Apple Silicon application?

If it's a native Intel application running under Rosetta, it could be the overhead caused by Rosetta.

I'd asked the OP this earlier, and he said that it's AS-native. The Gigperformer site confirms this:
"Gig Performer 4 is supported on Windows 11 and is available natively for Apple Silicon (new M1 and M2 chips)."

https://gigperformer.com/still-on-gig-performer-3-key-reasons-to-upgrade/

 

[theorist9]

ahh, not sure if i misspelled something ?
Gigperformer (GP) is a plugin Host. i run plugins within it.
GP can´t run as a plugin on itself.
There are some other hosts which can do that. GP can not.

Ah, got it. And you didn't misspell anything--you did say earlier that it's a plugin Host.

You mentioned some of your plugins aren't AS-native. Does most of the CPU load come from the plugins or the host? If it's from the plugins, could the Rosetta translation for the non-AS-native plugins be causing the performance issues? Do you have the issues even when using only native AS plugins?

Based on this (https://community.gigperformer.com/t/which-type-of-cpus-perform-best-with-gig-performer-4/9944/4), it does look like it's the plugins that load the CPU, so perhaps finding more efficient plugins (or at least ensuring all your plugins are AS-native) would help:

"For instance I’ve found the P-A Oberheim has some patches that can just jam the CPU utilization right through 100. I’m not exactly sure why this happens yet; and I’m not even sure if the Oberheim VST itself is spiking that—and I haven’t ruled out something else reacting in my processing chain. The Knifonium will do the same for some patches, but the Oberheim seems to have more such patches. It’s probably related to the built-in effects in the stock patch I would guess."

Also, later in that thread, it's mentioned that some plugins are multi-core, so it might be worth looking into those as well.

 

[TracerAnalog]

Hi there,

in several Max-Tech Vids vs. Benchmarking the new M3s,
it became directly visible that these M3 macs do not necessarily run at 4.05Ghz speed.

I´d like to know more about it.
Whats going on ? is this normal ?
When would these M3 macs go to full clock speed, and when not ?

here is a YT-video for example, check from 9:16 on.

personally i´m concerned about my music making.
My main app runs only ST.
The task is realtime-playing.
The CPU load is for example also determed by how many notes i play.
So, the whole CPU load i do generate is something very fluid !
I run my M2pro vs. my ST task into its knees.
I *really* need SC speed !

do i have to be concerned ?
Is this the same with the M2pros ?
How can i look the CPU speed of my cores in realtime ? What app do i need to have ?

So many question marks…
You say music making and ST task? What? I run Logic Pro with dozens of tracks, virtual synths, samplers, midi tracks, effects plugins and I never get it close to the red line (M1 Max, 64Gb). RAM is important for Samplers btw.
CPU frequency (solely) is just not something music studios focus on. It’s RAM and it’s acces times, latency (very much depends on your audio interface and driver), system stability, SSD read and write, compatibility with outboard gear etc.

BTW The number of notes I play is usually not more than 10 😉, which doesn’t even register on my CPU.

Also: MaxTech?😎

In short: Don’t focus on CPU frequency. Check some of the many DAW on Apple Silicon reviewers on YouTube.

/edit: ah you use GigPerformer, that explains your question. Disregard my answer 😀. Did you ever try Apple MainStage? Does it compare?

 

[streetfunk]

Did you ever try Apple MainStage? Does it compare?

No Mainstage is not a comparison.
Specific ? it´s a thing of who wants to do what exactly ?
So, it is finally usecase dependend to some degree.
But many (live gigging) folks move away from mainstage, over to GP.
for my specific uses is Mainstage just not anything interesting.
Overall ? GP wins by a big shot.

You mentioned some of your plugins aren't AS-native. Does most of the CPU load come from the plugins or the host?

The host is fine. I can run it under rosetta2 or ARM native.
I have full control over that ! And i do both.

I´ve made tests. Running the host ARM native is mostly something like 7-8% less CPU load vs. my patches.
Sometimes it´s more. Up to 15% max, iirc. But average its a ~7% win.
The host itself eats nearly nothing.

My high CPU load comes from running *many* FX plugins at once.
I also use sometimes *very* CPU hungry FX.
And i create very complex patches.

Overall "is it me" creating all that CPU load !
No, there is nothing i could optimise ! I´ve donne that allready.
Its the contrary, i could start to incorporate more oversampling to single plugins and could burn that way, way more CPU cycles.

Also, later in that thread, it's mentioned that some plugins are multi-core, so it might be worth looking into those as well.

I´m aware of that.
this is a own specific topic, and looks in real live a bit more complex than on paper.
Running single plugins multicore makes the whole thing mostoften more prone for audio crackling for example.

While gigPerformer runs initself only on one core, is it not hindering other plugins to run multicore on their own.
But thats a feature mostly used within plugin-instruments. And plays the more a role the more notes you´d play at the same time, and the longer you´d hold these. I do not use such instruments, and i do play mostly percussive.
The FX plugins i use are non multicore. Much likely all of them.
So, thats not exactly a backdoor for me, to lighten up the CPU load

"For instance I’ve found the P-A Oberheim has some patches that can just jam the CPU utilization right through 100. I’m not exactly sure why this happens yet; and I’m not even sure if the Oberheim VST itself is spiking that—and I haven’t ruled out something else reacting in my processing chain. The Knifonium will do the same for some patches, but the Oberheim seems to have more such patches. It’s probably related to the built-in effects in the stock patch I would guess."

ahh, you make music too ?
Yeah , such synths exist. Check "Kult" from "Dawesome".
Same as you described.

Unrelated, but what do you do exactly? Sounds cool!

you could call it realtime-play sounddesign what i am doing.
Its mostly realtime-play patches with the intention to make electronic hand-played instruments as expressively playable as real instruments are. Strings you can bend. You touch them, you vary the soundquality on the get go.
My work turns plugin instruments into "really playable" ones.
It´s so far an endless development process.......

We have the single technology bits ,-it takes to put them together-, available since something like 2.5 years.
Together with the AS macs, thats entirely a breakthru, imho.

i´m btw. never above ~32W, what my M2proMini is sucking on the mains while doing my music

 

[OldStrat]

personally i´m concerned about my music making.

Is there any place that we can check out your music?

I *really* need SC speed !

You can use multiple instances of Gig Performer. Starting from v4.7, you can also easily synchronize these instances. There are also ways to route audio/MIDI between these instances.

The first responder got an M2 Max with 96GB memory and says the machine is more powerful than he needs.

In reality, Gig Performer takes almost no system resources by itself. You can run it on crappy 13+ years old Windows computers.

Some FX have to be loaded into another host first,

Why don't you use Gig Performer's Auto Sampler feature? You can sample many CPU intensive patches and throw them in e.g. Decent Sampler, that is demonstrated in this video.

Additionally, there are many tips in their community on how to optimize your setup and tax your CPU less.

(I'm not sure how to use it actually!)

It's pretty easy - you simply connect plugins and other blocks using virtual wires.