Finding Bitwig Studio 1.3.15 more CPU efficient than Live! :-)

[humanbeingbeing]

Bitwig also has two seperate processes.
A fair comparison would need to be worked out no?

Yes Im going to do a few personal tests to ascertain the situation... I think itll be fairly clear which is performing well at what, I have some assumptions but I havent looked at numbers, in my experience Bitwig simply bogs down hard in comparable production scenarios, the ceiling seems lower but itll be an interesting discussion, looking at the Bitwig processes and whether its dropping out would have to factored in.

The point at which either drops out is what interests me.

[Regnas]

For the people saying that Live is CPU efficient, just fire up Live and Bitwig (or any other DAW) at same time, both with empty project, and then open the task manager and check the CPU usage of both....

Cheers...

Not a valid experiment. The initial overhead is not an indicator of cpu efficiency. I will try some experiments on the latest Bitwig this wkd if I get some free time, ill post the results here..

I'm not interested on the initial overload either, the idle part is what scares me... Any software eating a lot of CPU when idle it's never a good indicator. Unfortunately for me FL is falling in this problem too, with the latest releases..... I hope they can fix it..

But hey, do the tests, I would love to see the results.... And be prepared, Live will lose...

[humanbeingbeing]

quick basic test with 1024 sample latency with 6 instances of Serum playing the same midi part in either host shows very comparable results, if you drop a 7th serum both apps start dropping out severely, at 6 live hovers around 21% cpu, at 6 bitwig hovers around 20% cpu BUT Bitwig is occasionally dropping out... this isn't a scientific experiment, but it shows an example where you need not max out your CPU and still drop out... each host has a certain level of overhead but they even out when you're reaching processing limits.

I tried a few variations on the latency and some more FX tests and it's the same story, so far they seem to be performing equally, Bitwig is dropping out ever so slightly earlier but only very slightly... it's performing well, better than I expected or had experienced in the past, but it's not more efficient than Live, very close to it, not more than.

[humanbeingbeing]

For the people saying that Live is CPU efficient, just fire up Live and Bitwig (or any other DAW) at same time, both with empty project, and then open the task manager and check the CPU usage of both....

Cheers...

Not a valid experiment. The initial overhead is not an indicator of cpu efficiency. I will try some experiments on the latest Bitwig this wkd if I get some free time, ill post the results here..

I'm not interested on the initial overload either, the idle part is what scares me... Any software eating a lot of CPU when idle it's never a good indicator. Unfortunately for me FL is falling in this problem too, with the latest releases..... I hope they can fix it..

But hey, do the tests, I would love to see the results.... And be prepared, Live will lose...

Initial overhead... and overhead is there for a reason, could be buffering, crash protection, overhead is often to prevent dropouts, many things, kinda too complicated, you need to actually process things with music in order to ascertain CPU utilization... but I don't really think this is worth arguing over... I did several tests and the programs were neck and neck... the CPU numbers are basically the same when audio is being processed, 20-21% on both apps, the idle CPU is not added, they're the same numbers.

[pekbro]

That's what I would expect, for me they seem pretty on par with each other for the most part. Live is far more stable though IME.

I like them both, that's really all I'm concerned with.

-Cheers

[Regnas]

@humanbeingbeing:

Thanks for the quick test info....
Well, it seems like Live doesn't like me...

Cheers..

[xbitz]

just a remark Asio4All with its own inner buffer a definitive must have for Bitwig

[pboy]

Very interesting thread, and thanks to HumanB for taking the time to do a comparative test!

Here's a little more detailed info on the projects I'm finding more CPU efficient in Bitwig;

I use 30+ instances of the Kontakt sampler with Kontakt set to render the "perfect" fidelity at off-line mixdown (bounce, render audio). Kontakt's engine is also set to not support multiple cores, which I set the host to handle. I use two send/return fx tracks with 2cAudio's B2 and Aether, these reverbs are also set to render at the highest resolution (I guess it's some oversampling procedure). Every Kontakt track has its own EQ and compressor (saved with nested Instrument in Bitwig, which IMHO is a very handy way to work fast with your self-designed channel-strips; because the given preset names then do pop up as the arrange track name, something that doesn't work if using containers rather than nested instruments). Some plugins on the master output too set to render higher than playback fidelity; FabFilter ProQ, Eventide UltraChannel, FabFilter ProL). And a lot of the samples luckily fit onto the system SSD, which I'd guess might be important (smoother streaming in Bitwig?)

For some reason, I don't technically understand, the playback audio stays crisp and clean in Bitwig while Live starts crackling after a dozen tracks and has me start freezing tracks (like I always do in Logic, but that's on a different machine/OS).

[humanbeingbeing]

just a remark Asio4All with its own inner buffer a definitive must have for Bitwig

are you suggesting Asio4All over your native Asio driver? normally Asio4All is to give Asio support for built in soundcards over something like MME.... I used an RME BabyFace Pro when testing, and I used the native RME Asio... I'd be surprised if it were worse than Asio4All.

[humanbeingbeing]

Very interesting thread, and thanks to HumanB for taking the time to do a comparative test!

Here's a little more detailed info on the projects I'm finding more CPU efficient in Bitwig;

I use 30+ instances of the Kontakt sampler with Kontakt set to render the "perfect" fidelity at off-line mixdown (bounce, render audio). Kontakt's engine is also set to not support multiple cores, which I set the host to handle. I use two send/return fx tracks with 2cAudio's B2 and Aether, these reverbs are also set to render at the highest resolution (I guess it's some oversampling procedure). Every Kontakt track has its own EQ and compressor (saved with nested Instrument in Bitwig, which IMHO is a very handy way to work fast with your self-designed channel-strips; because the given preset names then do pop up as the arrange track name, something that doesn't work if using containers rather than nested instruments). Some plugins on the master output too set to render higher than playback fidelity; FabFilter ProQ, Eventide UltraChannel, FabFilter ProL). And a lot of the samples luckily fit onto the system SSD, which I'd guess might be important (smoother streaming in Bitwig?)

For some reason, I don't technically understand, the playback audio stays crisp and clean in Bitwig while Live starts crackling after a dozen tracks and has me start freezing tracks (like I always do in Logic, but that's on a different machine/OS).

Kontakt is not a particularly CPU heavy app, but 30+ could hit the CPU on a modern machine, I'd be curious if it were maybe the distribution of less resource intensive things... if you could share your projects I could have a look and see exactly what it is, but I have those plugins, I can do something similar on my home machine and see what result I get.. actually I'll fire it up now and see. I'm not going to include internal plugins because the eq and compressor and only roughly comparable between Bitwig and Live, I personally think they're higher quality in Live and I wouldn't be surprised if they use more cpu, we should know with the 30 kontakt samplers alone.

[humanbeingbeing]

Just did the same test roughly, 30 Kontakts at perfect no cores, instance of Aether, instance of B2, FabFilter Pro q and Pro L on the master (haven't got eventide)... same results as before, basically no difference between the two crucially they don't even hit the CPU, low numbers, 9-10% equally... there is crackling in both hosts but it's not due to the hosts, it's due to Kontakt disk streaming... unfortunately this is a bad real world test, asking Kontakt to stream the same sample from the disk is overloading the file system, even on a fast SSD... and this is a skewed test because a better performing app could actually have more disk drop outs due to tighter timing.

Long story short- Nothing to do with CPU, hosts perform equally, more of a Kontakt/user problem than a host issue.

ABLETON.png

BITWIG.png

[xbitz]

just a remark Asio4All with its own inner buffer a definitive must have for Bitwig

are you suggesting Asio4All over your native Asio driver? normally Asio4All is to give Asio support for built in soundcards over something like MME.... I used an RME BabyFace Pro when testing, and I used the native RME Asio... I'd be surprised if it were worse than Asio4All.

my Focusrite soundcard has only 10ms buffer with its own driver, ASIO4All has 46.44ms(442 samples vs 2048 samples), for me the CPU resource is more important than a little bit latency at the beginning(which can be reduced if I want to record something live), with 10ms a quite simpole layered pad(long release, long reverb decay time) timbre can kill Bitwig

[humanbeingbeing]

just a remark Asio4All with its own inner buffer a definitive must have for Bitwig

are you suggesting Asio4All over your native Asio driver? normally Asio4All is to give Asio support for built in soundcards over something like MME.... I used an RME BabyFace Pro when testing, and I used the native RME Asio... I'd be surprised if it were worse than Asio4All.

my Focusrite soundcard has only 10ms buffer with its own driver, ASIO4All has 46.44ms(442 samples vs 2048 samples), for me the CPU resource is more important than a little bit latency at the beginning(which can be reduced if I want to record something live)

so would Asio4All perform better if you could set both drivers to the same latency settings? this can't be true for all soundcards, I'm not sure you can suggest this over native drivers., it would be very quirky indeed if Asio4All performed better at the same settings.

[linkstate]

Live does have one of the better sample rate conversion algos, maybe thats why it uses more CPU

http://src.infinitewave.ca/

What is Ableton 9.11? :-O

[ThomasHelzle]

Guys, I have a question:
Some of you know that I always said that running plugins in the audio engine should be faster than running each one in it's own sandbox. (Plugin Preferences -> "only as bitbridge" vs. "every plugin"). It's a long time I did tests but recently I re-did an old test with multiple instances of Diva with the BWS 2 beta.
To my bafflement I found, that I can actually have twice as many tracks (10) with Diva playing 5 notes and a heavy preset with "every plugin" than with "only as bitbridge" (~5).

Could you guys maybe test your projects as well with the two different settings?
Make sure you close and restart Bitwig after changing the setting, you can't switch this on the fly.

Here is my testproject with Diva, that I can just about play without crackles with 48kHz/128 samples on an i7 6core @ 4.1 gHz:
http://www.screendream.de/stuff/Bitwig/ ... .bwproject

I asked the devs on the betalist about this massive difference and hope we can get to the bottom of this.

Cheers,

Tom