Bitwig drops the CPU Bomb!

[Machinesworking]

So, in a thread on another forum about various DAWs and CPU usage I was prompted to run a performance test again. I'll come back to this thread with Lives results, but we were talking about Logic, Digital Performer, Reaper and Bitwig.
Running a simple test of the INIT patch Diva instance playing an eighth note run for four measures cycled.
OSX Mojave, 2009 Mac Pro 3.33ghz 12 cores, at a 256 buffer in all four DAWs.
This is a classic "to failure" test. You add tracks until the audio starts messing up then back off.

Logic 10 instances
DP 11
Reaper 12
Bitwig 25

Live has always performed worse than Logic every time I've ever done this test in the past, and unless something has changed it will be the same.

Bitwig just killed everyone, literally wipes them off the map. They talked about improved performance but I didn't expect this! In the Activitly monitor in OSX it shows Bitwig running a separate process to the OS for every instance of the plug in, thereby giving the OS the task of allocating resources for the plug in, and just destroying all other DAWs in the process. I have no idea why they aren't shouting this from the rooftops?

For the first time since Bitwig came out I'm seriously thinking of switching.

[edited for spelling]

[8E]

You are right, there is no more excuse for CPU spikes and hot computers in our computing age.

Still, I need Live.

[Machinesworking]

It’s specifically Bitwig that’s getting fantastic performance though. It used to score the same as Live, which scores significantly lower than Logic, Cubase, DP etc.

They did something pretty amazing really.

[siliconarc]

could you share the forum link pls? interested to take a look

[jonljacobi]

I’d be careful about this. Bitwig’s audio engine still has some issues syncing warped with nonwarped audio. This stuff is hardly rocket science at this point. It’s well understood and for any program to double the performance of the others is very odd.

[Machinesworking]

Pardon me for being very skeptic here, up to a point where I'm having a hard time taking this serious (no offense).

See: stress tests seriously don't mean all that much within this context. They're good to put load onto an engine to see how the engine behaves under stress, but that usually doesn't even come close to resembling a real working environment.

And considering the way Bitwig started and all... I wouldn't even be surprised if they have added routines in their software in order to cope with such stresstests and address the testing environment directly. In other words: ensuring that the test routine gives a much better result because the software itself caters to providing a higher priority in order to get those better results.

They wouldn't be the first to do this.... which also adds up to my skepticism... I don't trust stress tests, at all. The only thing it's good for is to put a load on your hardware to see how well it comes (like an SSD to get a true speed readout) but software? That's not reliable at all.

I respectfully disagree. Over the course of 15 years, using and testing at least 6 different DAWs on roughly ten machines, stress tests are a great indicator of real world performance. Live for instance consistently gets roughly 70% of the performance CPU wise of Logic, Reaper, or DP on my system. This is without any time stretching or other CPU intensive task, just virtual instruments and CPU.

Now if you're thinking about Apple showing off Logic with their latest hardware with Logics built in instruments, yeah, that's cheating.

In the test I did Diva was used, it's not made by Bitwig, and it's clearly running as a process in separate instances in OS X, Bitwig are using the systems core allocation to get better performance, and the shocking thing to me is it's an old as the hills Mac Pro, 2009 with a modified chip base of 12 3.33Ghz cores. I updated to the latest version of Bitwig today and got 36 instances! !

Here's the weird thing though, it's picky about machines, and doesn't like the Macbook Pro 2012 2.7ghz 4 core i7 here, arguably a better chip scoring much higher in Geekbench than the Xeons in the Mac Pro? I mean it really doesn't like it, I'll get the numbers VS Live 10 Suite in a second here.

[Machinesworking]

Because this stuff interests me, and I'm curious here are the results for the Mac Pro.
The Macbook Pro on initial tests is showing really different results, so I'm curious about whether this is CPU or Graphics card specific in terms of Bitwig?

09 Mac Pro 12 core 3.33Ghz Xeon, 24GB RAM, Radeon RX 580 8 GB graphics, fast M.2 SSD for the boot drive, OS X Mojave latest version.
Simple eighth note four bar loop with U-He Diva

Live 10 - 9 instances.

Logic X - 10

DP10 - 11

Bitwig - 36 <--- This is insane, and they are all running in separate processes on the OS in in activity monitor. On this machine it really helps, but wait for it! On the i7 2.7 Macbook Pro the tests will be vastly different. (no Logic though, don't have it installed there)

[Artcutech]

I'll be honest with you my friend, this sound amazing. Actually thinking of downloading the demo sometime soon after reading this and getting familiar with it, what would you say is the hardest part getting use to when switching from Ableton?

[Machinesworking]

I'll be honest with you my friend, this sound amazing. Actually thinking of downloading the demo sometime soon after reading this and getting familiar with it, what would you say is the hardest part getting use to when switching from Ableton?

I can't really tell you, I haven't switched yet. But one thing is very certain, Bitwig did some great things with CPU allocation in the last year or so, and for whatever reason aren't bragging about it. Might be some bugs with some people's systems you never know?

I know one thing, Bitwig loves with a passion this old modified 12 CPU Mac Pro here. No other DAW on my system is using all 24 possible threads, but Bitwig is (with super minor hiccups here and there adjusting past 28 instances or so) using all 24. You can clearly see it in Activity monitor. You can also hear it. I just played 36 instances of Diva with one that I was banging on a keyboard on, with no real stuttering.

Compare this to Live, at 9 instances it's done, and clearly not multithreading the cores.

I think getting over how busy it looks is my major hurdle, it doesn't usually take me too long to learn the basics of a DAW, except Reaper which was a PITA to set up.

I also don't think it's Clip launcher is as advanced as Lives? I could be wrong though. I'm definitely thinking of picking it up though, the CPU performance on this old Mac makes it worth it alone.

[Machinesworking]

To be clear, I'm a skeptic as well, but the numbers don't lie.
Bitwig are doing proper multithreading on both my old Mac Pro and old Macbook Pro.
I'm certain the numbers wouldn't be surprisingly different for more modern macs or PCs.
These numbers represent single instances of Diva in INIT mode, playing an eighth note four bar loop at 120 BPM with a 256 buffer in each DAW.

09 Mac Pro 12 core 3.33 Xeon CPU use

Bitwig - 36

Live 10 - 9

Logic - 10

DP10 - 11

Reaper - 11


2012 4 core 2.7 i7 Macbook Pro

Bitwig - 14

Live10 - 7

DP10 - 9

Reaper - 15

[8E]

As you have noticed, the CPU speed is not any more going up as much as before. We have reached a long time ago CPU at 3GHz, and today's computers don't need more.

OTOH, applications are developed so much forward that many of them can be run on a TABLET that doesn't have fan or cooling.

This is very very important: we need computing engine that is more efficient, without need to have a computer bomb of 3THz nor 16PB RAM in order to run Ableton "smoothly", neither any other app.

Long time ago I have noticed that my computer gets insane hot with AL, and running the same thing in Logic makes it much cooler.

[Machinesworking]

As you have noticed, the CPU speed is not any more going up as much as before. We have reached a long time ago CPU at 3GHz, and today's computers don't need more.

Well kind of, CPUs are getting faster, but the clock speed isn't the only number that's important.

Look at the tests, the 2.7 four core i7 is calculating 3.5 instances of Diva per core, while the "faster" 3.33 ghz 12 core is calculating 3 instances of diva per core. Some of this may be down to the inefficiency of multi core chips VS less cores, but it's also due to different architecture, different floating point etc. speeds.


With Live and the others you get varying degrees it seems support for multithreading and the two chips. Bitwig seems to be the most consistent, the others are using the i7 slightly better than the Mac Pro, some only run 12 threads out of 24 on the Pro, but are running the possible 8 on the macbook pro.

[jonljacobi]

I’d be careful about this. Bitwig’s audio engine still has some issues syncing warped with nonwarped audio. This stuff is hardly rocket science at this point. It’s well understood and for any program to double the performance of the others is very odd.

That said, the vast majority of the performance improvements these days come from multiple cores processing in parallel, and Bitwig does seem more modern in this regard. But my older computers far exceed my needs and Bitwig has some real limitations when it comes to the basics, so this is basically interesting trivia in my case.

[jlgrimes]

Last comment from me:

To be clear, I'm a skeptic as well, but the numbers don't lie.

Unless you actually know how Bitwig is programmed you cannot draw such conclusions, I explained my reasoning above. Sorry, but blindly trusting numbers as shown here (especially such drastic ones) makes it very hard to take your comment about being a skeptic seriously.

I'm sure that test dont tell the whole story. But the numbers are still impressive and at least when it comes to Diva, Bitwig is winning. But these days I'm not even sure if Diva is the biggest resource hog anymore. Certain patches in particular aren't that much different than typical softsynths.

It is still hoggy but Repro 5 seems more hoggy and programs like Massive X, and Pigments can be hoggy as well now.

[Machinesworking]

Last comment from me:

To be clear, I'm a skeptic as well, but the numbers don't lie.

Unless you actually know how Bitwig is programmed you cannot draw such conclusions, I explained my reasoning above. Sorry, but blindly trusting numbers as shown here (especially such drastic ones) makes it very hard to take your comment about being a skeptic seriously.

It's patently easy to see what's happening. With the 24 process's available in in the Mac Pro 12 core on OS X, Bitwig is using 95-95% of all of them, evenly. OS X has an Activity Monitor it shows CPU usage over time for all available threads for the computer you have, it's 24 threads on my machine.
Guess what Live does? it uses only one of the two possible threads a core can use, in Activity Monitor every other thread is empty, and each CPU is only at about half filled before Live seems to bog down. Literally, OS X's own CPU monitoring system shows Live crapping out at 1/2 the performance on the modern i7 that it uses 8 threads on, and 12 on the Xeon in the Mac Pro. It's also the only DAW that's reporting each track with a plug in on it as a separate process to the OS, this is clearly visible and again seems to be part of how they're getting such good results. This all seems to be something they started around

The OS itself backs up the performance differences between the DAWs, it's not just smoke and mirrors as you're suggesting. Take it from someone who doesn't currently own Bitwig, thinks it's ugly and wasn't that happy initially with the direction they were going in. They seem to have drastically changed the way plug ins are hosted around 2.5 and it's pretty refined at this point.