Page 1 of 2

Audio interface and CPU Load

Posted: Sat May 08, 2021 8:41 am
by WilmarBoer
I'm just curious if there's a difference in CPU load and least possible latency between audio interfaces on the mac using audiocore.

I have a Behringer UMC1820. Happy with the possibilities and sonic results of the interface. But at 8 tracks and just freezing MIDI tracks still gives it a 25% CPU load with a latency of 256ms. Let alone a project with 16 or more tracks, VSTi's and quite a few FX. I'm on a Macbook 2019 - 16 inch, i9, 8-core, 2,3 ghz with 16gb RAM. So my macbook is supposed to be quite fast. I really like to go down to 64ms to give vocalists and myself as a guitar and keys player the best experience of playing a track in good timing using Ableton FX (in stead of using external FX through direct monitoring).

Before I put my money into a new audio interface. Any experiences on tackling this issue by swapping out the audio interface?

Re: Audio interface and CPU Load

Posted: Sat May 08, 2021 10:09 am
by Dallon426
Absolutely.

The best interfaces in the business for stability and latency are RME interfaces. Period.
You cannot get better support across multiple OS and they have the lowest possible RTL available.
The thing about buying interfaces is that you need confidence and less hastle. Rme is still writing drivers for ten year old interfaces. Where other companies just release "new" interfaces and stop support for the old ones.

Your interface is crucial for audio production. One of the most important pieces. Can you make music with cheaper interfaces... Yes. But most of the time you'll be frustrated because of the constant issues and crackle and pops. I do not experience that with my RME Babyface pro.

Find a used one and save some cash.

Re: Audio interface and CPU Load

Posted: Sat May 08, 2021 10:34 am
by miyaru
I agree, RME has the best RTL of all, and is an excellent interface. But if the price is to hight and you need as standand more inputs a Focusrite Scarlet 18i20 3rd gen is a good alternative.

I run a Win 10 rig with an i7-7700 quad core, 16 Gb ram and a NVMe drive and go down to 64 samples buffersize. I can run like 25 tracks like this in Live 10 suite. At 128 samples it is getting much better, but it runs normally at 64 samples.

At less then 500 dollar this is a good interface! And yes I own one.....

Re: Audio interface and CPU Load

Posted: Sat May 08, 2021 7:05 pm
by Mark Williams
I have the 2nd gen 18i20 scarlett and get similar results, good interface for the money :)

Re: Audio interface and CPU Load

Posted: Sat May 08, 2021 7:17 pm
by Dallon426
It depends, what is the latency on them? I am a pianist and guitarist so for me latency is what I am after. BTW your focusrite interfaces are one behind. Meaning if they are saying it is 32 buffers, it really is 64, and if it says 64 it is really 128, so on and so forth

RTL for Focusrite scarlett
032 (016)* 4.150**
064 (032)* 4.680**
128 (064) 8.037**
256 (128) 12.955**
512 (256) 24.863**

You can view the latency results here

http://dawbench.com/images/DAWbench%20L ... 202021.pdf

Re: Audio interface and CPU Load

Posted: Sat May 08, 2021 8:09 pm
by Mark Williams
latency in my PC is 9.2ms @128 samples 48hz so im happy as a semi pro drummer low latency extremely important... Mine will go down to 32 samples easily, but Im happy @128

Re: Audio interface and CPU Load

Posted: Sun May 09, 2021 11:37 am
by WilmarBoer
Dallon426 wrote:
Sat May 08, 2021 7:17 pm
It depends, what is the latency on them? I am a pianist and guitarist so for me latency is what I am after. BTW your focusrite interfaces are one behind. Meaning if they are saying it is 32 buffers, it really is 64, and if it says 64 it is really 128, so on and so forth

RTL for Focusrite scarlett
032 (016)* 4.150**
064 (032)* 4.680**
128 (064) 8.037**
256 (128) 12.955**
512 (256) 24.863**

You can view the latency results here

http://dawbench.com/images/DAWbench%20L ... 202021.pdf
Thanks! But I guess the PDF document refers to a Windows install with windows drivers made by the interface manufacturers brand. While on a Mac Apple Audiocore is 'talking' to the interface and as long as the interface is audiocore compliant it will work without driver install. But then Audiocore will be of influence in these latency figures.

And interesting to see that ESI is quite fast. Actually quite good equipment without beïn expensive.

Re: Audio interface and CPU Load

Posted: Sun May 09, 2021 11:38 am
by WilmarBoer
Dallon426 wrote:
Sat May 08, 2021 10:09 am
Absolutely.

The best interfaces in the business for stability and latency are RME interfaces. Period.
You cannot get better support across multiple OS and they have the lowest possible RTL available.
The thing about buying interfaces is that you need confidence and less hastle. Rme is still writing drivers for ten year old interfaces. Where other companies just release "new" interfaces and stop support for the old ones.

Your interface is crucial for audio production. One of the most important pieces. Can you make music with cheaper interfaces... Yes. But most of the time you'll be frustrated because of the constant issues and crackle and pops. I do not experience that with my RME Babyface pro.

Find a used one and save some cash.
Thanks! I'm looking into that one. In the video editing suite in our business we work with RME also.

Re: Audio interface and CPU Load

Posted: Sun May 09, 2021 11:44 am
by Dallon426
WilmarBoer wrote:
Sun May 09, 2021 11:37 am
Dallon426 wrote:
Sat May 08, 2021 7:17 pm
It depends, what is the latency on them? I am a pianist and guitarist so for me latency is what I am after. BTW your focusrite interfaces are one behind. Meaning if they are saying it is 32 buffers, it really is 64, and if it says 64 it is really 128, so on and so forth

RTL for Focusrite scarlett
032 (016)* 4.150**
064 (032)* 4.680**
128 (064) 8.037**
256 (128) 12.955**
512 (256) 24.863**

You can view the latency results here

http://dawbench.com/images/DAWbench%20L ... 202021.pdf
Thanks! But I guess the PDF document refers to a Windows install with windows drivers made by the interface manufacturers brand. While on a Mac Apple Audiocore is 'talking' to the interface and as long as the interface is audiocore compliant it will work without driver install. But then Audiocore will be of influence in these latency figures.

And interesting to see that ESI is quite fast. Actually quite good equipment without beïn expensive.
RTL is the same more or less in both windows and mac.
I have done the same tests with windows and mac and have seen no significant difference. Windows uses asio and mac applecore. It doesn't make a big difference in the grand scheme of things.
I just tested the RME Babyface pro on mac yesterday and got similar results as my windows build.

Re: Audio interface and CPU Load

Posted: Sun May 09, 2021 11:45 am
by WilmarBoer
Dallon426 wrote:
Sun May 09, 2021 11:44 am
WilmarBoer wrote:
Sun May 09, 2021 11:37 am
Dallon426 wrote:
Sat May 08, 2021 7:17 pm
It depends, what is the latency on them? I am a pianist and guitarist so for me latency is what I am after. BTW your focusrite interfaces are one behind. Meaning if they are saying it is 32 buffers, it really is 64, and if it says 64 it is really 128, so on and so forth

RTL for Focusrite scarlett
032 (016)* 4.150**
064 (032)* 4.680**
128 (064) 8.037**
256 (128) 12.955**
512 (256) 24.863**

You can view the latency results here

http://dawbench.com/images/DAWbench%20L ... 202021.pdf
Thanks! But I guess the PDF document refers to a Windows install with windows drivers made by the interface manufacturers brand. While on a Mac Apple Audiocore is 'talking' to the interface and as long as the interface is audiocore compliant it will work without driver install. But then Audiocore will be of influence in these latency figures.

And interesting to see that ESI is quite fast. Actually quite good equipment without beïn expensive.
RTL is the same more or less in both windows and mac.
I have done the same tests with windows and mac and have seen no significant difference. Windows uses asio and mac applecore. It doesn't make a big difference in the grand scheme of things.
I just tested the RME Babyface pro on mac yesterday and got similar results as my windows build.
Thanks!

Re: Audio interface and CPU Load

Posted: Fri May 14, 2021 1:02 pm
by WilmarBoer
I tried using the Mac internal audio and a ZOOM U44 interface. With no difference. As soon as I go below 256 samples the audio becomes unstable and CPU is often maxed-out.

Re: Audio interface and CPU Load

Posted: Fri May 14, 2021 2:18 pm
by Calagan
My 2 cents : I bough recently a little MOTU M2 to replace temporarily a Focusrite Clarett (thunderbolt) that was on repair, and I was very surprised to actually find it sounds better, has less latency and has a bit less impact on the CPU.

On my macbook pro at least (mid-2012 running Mojave) and after I installed the Mac drivers (it's not the same story on Windows) it's super stable, very low latency and it sounds truly amazing. To me, better than a 2 times more expensive interface like the Clarett.

Here are the latency charts I see in Ableton live @44kHz (I didn't test it scientifically)
64 samples = 4ms RTL
128 samples = 6.89ms RTL
256 samples = 12.7ms RTL

To me, it's the best interface I ever had regarding sound quality, latency and stability/CPU efficiency, and it's the cheapest... Go figure...

Re: Audio interface and CPU Load

Posted: Fri May 14, 2021 2:51 pm
by WilmarBoer
Nice! BUt is that with a mix running? The song in particular where I'm working on is about 24 tracks. All with simple plugins like compression and EQ and global Reverb. I've already freezed the VSTi channels.

Re: Audio interface and CPU Load

Posted: Mon May 17, 2021 12:54 pm
by jlgrimes
Calagan wrote:
Fri May 14, 2021 2:18 pm
My 2 cents : I bough recently a little MOTU M2 to replace temporarily a Focusrite Clarett (thunderbolt) that was on repair, and I was very surprised to actually find it sounds better, has less latency and has a bit less impact on the CPU.

On my macbook pro at least (mid-2012 running Mojave) and after I installed the Mac drivers (it's not the same story on Windows) it's super stable, very low latency and it sounds truly amazing. To me, better than a 2 times more expensive interface like the Clarett.

Here are the latency charts I see in Ableton live @44kHz (I didn't test it scientifically)
64 samples = 4ms RTL
128 samples = 6.89ms RTL
256 samples = 12.7ms RTL

To me, it's the best interface I ever had regarding sound quality, latency and stability/CPU efficiency, and it's the cheapest... Go figure...
Sounds impressive, I would need to research.

I'm currently using a Scarlett 3rd gen which works pretty good but from going from Apple on a UAD apollo to Windows on a much higher spec'd machine (although with no Thunderbolt), I went with Scarlett. I noticed that the Mac seem to have lower CPU loads (using Ableton CPU meter or I think it did), but the main thing I do notice is MultiClient audio performance was much better on Mac. I'm guessing its probably the Thunderbolt aspect helping here.


My RTL is almost twice your numbers using the Scarlett.

That said the Apollo's RTL numbers. weren't that impressive as well (but that had something to do with how you could use the UAD plugins at input with no latency).

Re: Audio interface and CPU Load

Posted: Tue May 18, 2021 2:54 pm
by kitekrazy
WilmarBoer wrote:
Sat May 08, 2021 8:41 am
I'm just curious if there's a difference in CPU load and least possible latency between audio interfaces on the mac using audiocore.

I have a Behringer UMC1820. Happy with the possibilities and sonic results of the interface. But at 8 tracks and just freezing MIDI tracks still gives it a 25% CPU load with a latency of 256ms. Let alone a project with 16 or more tracks, VSTi's and quite a few FX. I'm on a Macbook 2019 - 16 inch, i9, 8-core, 2,3 ghz with 16gb RAM. So my macbook is supposed to be quite fast. I really like to go down to 64ms to give vocalists and myself as a guitar and keys player the best experience of playing a track in good timing using Ableton FX (in stead of using external FX through direct monitoring).

Before I put my money into a new audio interface. Any experiences on tackling this issue by swapping out the audio interface?
That's why. CPU isn't string enough.