DSP reduce idea

[Sojuzstudio]

It occurred to me to try with Jack in all 3 OS's - but it's been a nightmare trying to make Jack run in Linux

I have Jack working directly from the package. Have you tried AvLinux?

[johne53]

Have you tried AvLinux?

Ah, we're back on the treadmill... 

There'a always some other distro where everything works 

For me, the easiest and most reliable distro was OpenSuse but I figured I'd try Zorin because it offers a "Windows like" feel (which to be fair, it does...)

[Sojuzstudio]

This is my home machine and not so powerful

Processor Intel® Core™ i5-6600K CPU @ 3.50GHz
Memory 32753MB (4525MB used)
Machine Type Desktop
Operating System Debian GNU/Linux 11 (bullseye)
Mixbus 32C 8.0.17
(rev 8.0-17-g39fc2c0f9f)
Intel 64-bit

"Love Having You Around"
All tracks
buffer 512
Jack Idle 21%
Jack play 21%

Alsa Idle 29%
Alsa play 33%

Same session on Ardour 6.9

buffer 512
Jack Idle 4%
Jack Play 5%

Alsa Idle 6%
Alsa Play 6%

[drsaamah]

oh man, i really don't start this argument. god bless, i hope everyone is using the OS that makes most happy
back to the topic at hand, i think @sojuzstudio has the numbers. i'm still on version 32Cv7 and i already uninstalled v6. however, like i said, that code i posted from the Ardour forums/github really did it. last night I was running a project with 52 tracks, 128 sample buffer size, 5 external sends, 2 external returns, plugins here and there and my DSP usage was hovering around 32% during playback. those of you that know how windows and/osx works have to look into that CPU DMA latency to find the solution i think. i feel like OSX users should be able to launch mixbus from terminal line to get some insight into what's going on. maybe windows users can do a similar thing (does windows still have dos prompt?).

[johne53]

Thanks Sojuzstudio, that kinda brings us full circle and illustrates the point that Tassy made in the first place... i.e. if you've a truly powerful machine (like me & Dingo seem to have) there isn't a massive difference between ver6 and ver8. But if you've a moderately powerful machine (though maybe not the absolute most powerful) each new release of MB brings a noticeable deterioration in DSP usage. And in some cases Jack gives a big improvement and other cases only a small improvement (or no improvement...)

I just wish we understood why... 

[Dingo]

This is my home machine and not so powerful

Processor Intel® Core™ i5-6600K CPU @ 3.50GHz
Memory 32753MB (4525MB used)
Machine Type Desktop
Operating System Debian GNU/Linux 11 (bullseye)
Mixbus 32C 8.0.17
(rev 8.0-17-g39fc2c0f9f)
Intel 64-bit

"Love Having You Around"
All tracks
buffer 512
Jack Idle 21%
Jack play 21%

Alsa Idle 29%
Alsa play 33%

Same session on Ardour 6.9

buffer 512
Jack Idle 4%
Jack Play 5%

Alsa Idle 6%
Alsa Play 6%

I would like to see your figures with a buffer of 1024 which is what I did.
I find 1024 is the optimum for mixing and that is what I am interested in testing.
Tracking is a different beast low latency = high cpu.
With this test if I go to 512 the cpu doubles yet at 2048 the cpu is no lower.

Interesting that JACK is getting better figures than ALSA...

Also comparing Ardour is a bit misleading ...
yes the engine behind Mixbus is Ardour, but Mixbus has a higher base CPU load because it has Mixer functionality baked in:
EQ, Dynamics , Busses, Saturation, Limiter etc that is 'optional' in Ardour.

So if you get the time could you post the specs at 1024 ?

[Dingo]

Thanks Sojuzstudio, that kinda brings us full circle and illustrates the point that Tassy made in the first place... i.e. if you've a truly powerful machine (like me & Dingo seem to have) there isn't a massive difference between ver6 and ver8. But if you've a moderately powerful machine (though maybe not the absolute most powerful) each new release of MB brings a noticeable deterioration in DSP usage. And in some cases Jack gives a big improvement and other cases only a small improvement (or no improvement...)

I just wish we understood why... 

Not so sure John,
I question how powerful my computer is... a 6 Core i5 macmini with 32GB RAM seems tame when others are talking 20 core Xeon etc.
I think there are so many combinations of interface, buffer size and the host capability that all come in to play.

If you do test , test using as many of the same parameters as you can.

[johne53]

Thanks Dingo - and I hadn't noticed that Sojuzstudio's comparison was with Ardour v6, rather than Mixbus v6 

[Sojuzstudio]

Ardour test, i was thinking to illustrate, how much Mixbus eats DSP on my machine. Sorry for confusion.

[Harvey.Stuart]

Thanks Sojuzstudio, that kinda brings us full circle and illustrates the point that Tassy made in the first place... i.e. if you've a truly powerful machine (like me & Dingo seem to have) there isn't a massive difference between ver6 and ver8. But if you've a moderately powerful machine (though maybe not the absolute most powerful) each new release of MB brings a noticeable deterioration in DSP usage. And in some cases Jack gives a big improvement and other cases only a small improvement (or no improvement...)

I just wish we understood why... 

Not so sure John,
I question how powerful my computer is... a 6 Core i5 macmini with 32GB RAM seems tame when others are talking 20 core Xeon etc.
I think there are so many combinations of interface, buffer size and the host capability that all come in to play.

If you do test , test using as many of the same parameters as you can.

Probably not the same machine and settings overall  but at least it's the same CPU (model name : Intel® Core™ i5-6600K CPU @ 3.50GHz), here are my numbers crunched on the same project using Mixbus32C8 and Alsa :

Code:

Mixbus32C8 (built using 8.0-17-g39fc2c0f9f and GCC version 6.3.0 20170516)
Alsalib v1.2.7.1

Import: 12s
DSP idle        12..13%
DSP play        13..14%
DSP idle        32..33% XTs
DSP play        43\33%  XTs