I just installed Linux on a Mac Mini and it looks (too?) good!

[kkrekula]

Hi,

as Macs nowadays run on pretty "standard" hardware, I installed Ubuntu Mate on my Mac Mini (and adapted the Ubuntu Studio packages).

After setting up dual boot, I tested the performance of Mixbus 3.6 both in Mac and Linux using the same song. Surprisingly there were big performance gains using Linux:

Mixbus DSP was running at about 36% in OSX, 23% in Linux and the system load was much lower.

I also performed an export test (exported the song to ogg vorbis) and Mixbus was 35% faster in Linux!

Has anyone of you experienced the same?

Regards
Kenneth

[Frank]

Hi,

as Macs nowadays run on pretty "standard" hardware, I installed Ubuntu Mate on my Mac Mini (and adapted the Ubuntu Studio packages).

After setting up dual boot, I tested the performance of Mixbus 3.6 both in Mac and Linux using the same song. Surprisingly there were big performance gains using Linux:

Mixbus DSP was running at about 36% in OSX, 23% in Linux and the system load was much lower.

I also performed an export test (exported the song to ogg vorbis) and Mixbus was 35% faster in Linux!

Has anyone of you experienced the same?

Regards
Kenneth

I have not. While testing i saw that Mixbus 3.7 was faster: Ben made me aware of the following qoute from the improvements :

Previous versions would only utilize physical processors (ignoring hyperthreading processors) This has been fixed to match operation on linux and windows.

It would be interesting if you could repeat your test with Mixbus 3.7 !

[kkrekula]

I have not. While testing i saw that Mixbus 3.7 was faster: Ben made me aware of the following qoute from the improvements :

Previous versions would only utilize physical processors (ignoring hyperthreading processors) This has been fixed to match operation on linux and windows.

It would be interesting if you could repeat your test with Mixbus 3.7 !

Ok I will do that as soon as you people here do not find any serious issues with v3.7 I am in the middle of a recording project so I don't want to screw up anything

/Kenneth

[Frank]

I have not. While testing i saw that Mixbus 3.7 was faster: Ben made me aware of the following qoute from the improvements :

Previous versions would only utilize physical processors (ignoring hyperthreading processors) This has been fixed to match operation on linux and windows.

It would be interesting if you could repeat your test with Mixbus 3.7 !

Ok I will do that as soon as you people here do not find any serious issues with v3.7 I am in the middle of a recording project so I don't want to screw up anything

/Kenneth

I understand your concern : but 3.7 was released because there were no serious issues.
But one can always install alongside each other. Rename the installed and install the next.

regards

[kkrekula]

I performed a comparison of v3.6 vs v3.7 on my Mixing PC (a quite new Intel I7 system) and the export test was 10 % faster with v3.7. The system and dsp load looked the same though.

Maybe because of the gcc5 version? (I have installed the gcc4 version previously, but this time the Mixbus installer warned about that, so I used the gcc5 version instead)

/Kenneth

[Jostein]

I have not. While testing i saw that Mixbus 3.7 was faster: Ben made me aware of the following qoute from the improvements :

Previous versions would only utilize physical processors (ignoring hyperthreading processors) This has been fixed to match operation on linux and windows.

It would be interesting if you could repeat your test with Mixbus 3.7 !

Does that mean that one should enable hyperthreading?

[Frank]

I have not. While testing i saw that Mixbus 3.7 was faster: Ben made me aware of the following qoute from the improvements :

Previous versions would only utilize physical processors (ignoring hyperthreading processors) This has been fixed to match operation on linux and windows.

It would be interesting if you could repeat your test with Mixbus 3.7 !

Does that mean that one should enable hyperthreading?

Yes : one should : and with my macs i cannot switch it off

regards

[Jostein]

Does that mean that one should enable hyperthreading?

Yes : one should : and with my macs i cannot switch it off

regards

I always thought that hyperthreading should be disabled when working with audio. Well, I enabled it...

...and what a RT-speed improvement! I just ran u-he's Repro-1 and Diva in diva mode plus ML-VST's PX8X (a JX-8P emulator for Windows) trough wine simultaneously with latency of 2.7 ms without xruns! Incredible! :-)

[Frank]

Does that mean that one should enable hyperthreading?

Yes : one should : and with my macs i cannot switch it off

regards

I always thought that hyperthreading should be disabled when working with audio. Well, I enabled it...

...and what a RT-speed improvement! I just ran u-he's Repro-1 and Diva in diva mode plus ML-VST's PX8X (a JX-8P emulator for Windows) trough wine simultaneously with latency of 2.7 ms without xruns! Incredible! :-)

Thanks for reporting !
I think hyperthreading is not efficient in pure DSP mode aka streams of data treated with the same instructions. But in a daw we do so many task switches that hyperthreading becomes of use again.
I get the feeling that the Mixbus code of 3.7 is also (more) optimized for hyperthreading.

although when running on a Mac i do not see more hyperthreading activity in the CPU history

regards

Continue testing Spectrum compressor

[BigstevE]

I too, always thought hyperthreading was a no-go at least in Linux audio. From memory, that comes from reading up on older (kernel 2.4-2.6) possible outdated Linux audio documentation.

The Spectrum Compressor:

Haven't downloaded 3.7 yet but based on the description all I can say is Wow !
Like a more intelligent Multiband Comp on steroids. Can't wait for a more detailed review, demos from early adopters on this one. Good Work Guys !