DAW Performance Test Results

[EgM]

I bought S1 about a week ago, through classifieds, got a good deal. I wasn't expecting to be as impressed with it as I am. I'm quite impressed with many aspects of it, including performance, but also workflow. I feel drawn towards using it and learning it. The devs are very responsive and are working really hard on it, it has a very bright future. One thing that stood out...S1 was able too use the VST3 version of the VEP plugin, which means it easily handled multiport to a single VEP instance, much like Cubase and DP can (and Logic can't without hacks and work arounds).

Yeah, lots happened since the 4.x release. The fast workflow alone made me sacrifice Cubase's advanced MIDI editing. Would've probably stayed in Cubase if it had the ability to focus VSTis to screen when you change tracks like Logic and Studio One does.

Cubase Pro 9.5 worked perfectly for me in terms of speed and stability, but the windows scattered everywhere, jittery display speed and subpar mixer was too much for me.

 

[Dewdman42]

Good test and good work! Thank you! I have many questions...

Will address some of your questions, but not all as some are irrelevant and distracting from the actual test. The point of the test was simply to compare a real world performance between several DAW's, attempting as much as possible to keep the test equal. Other tricks to increase performance or analyze specific aspects of performance such as particular sound cards, etc..are a test for another day.

• How many VEP instances were active? Connected?

One VEP instance in all cases. Note that for LogicPro I had to use my multiport macro templates in order to accomplish that. And notably, it performed very well.

• How many VEP audio channels per instance? MIDI ports?

One VEP instance in all cases, it used 8 midi ports. There were 90 channels. I can share some project files if people are interested, but because I used all VSL instruments, it will not be interesting for the vast majority of people. A better test would be to use freely available instruments of some kind in a methodical test that anybody can run on any system, but unfortunately I don't have that as of this time. But anyone else could use the same approach I used and try such a thing and I look forward to seeing the results!

• How many cores per instance of VEP?

On my 12 core system I had VEP configured for 18 threads per instance, with a single VEP instance. Could have probably set it to 20, but main point was consistency between each DAW..they all used the same configuration of one instance and 18 threads per instance in VEP.

• How many VEP buffers per instance?

The VEP buffer setting in the plugin only effects live operation, it does not effect playback. In any case I had it set to either 1 or 2, can't remember now, but it was the same for all tests.

• VEP on the same system or a slave?

Same system

• Were the DAW sessions exactly 90 tracks each?

yes exactly same midi tracks were used. The specific project I used is the E.T. score mockup that is available for free download from VSL's website. You need a lot of VSL products to use it though. In all cases, its 90 tracks of midi, feeding 91 instruments (the apassionata and orchestral basses are actually doubled from the same midi track). Each channel goes into MirPro and there is an instance of Miracle on the master bus, Also room tone added. Kind of a typical setup. Its the same configuration in each DAW, same configuration when hosted with or without VEP, the only difference between that without VEP, the DAW is hosting 91 ViPro instruments with MirPro and miracle rather then having those hosted in VEP.

As I said, the point was to have a consistent scenario between each DAW in order to compare them all essentially playing the same thing.

A perhaps more universal version of this test would be to use some freely available instruments and plugins and build up something that just throws a long stream of 16th notes or 32nd notes or something that will impact the CPU in a manner similar as a real project would do..then create that project on each DAW with and without VEP and measure it the way I did. Then everyone could run that on their system at home to compare one DAW to another, but of course...my results can't be compared to your results. we have different hardware. I can only compare my own results against my own results of different DAW's...and other people would need to do the same thing to have meaningful analysis.

 

[rgames]

Interesting. But 1024 is a pretty large buffer - what VI Pro buffer did you have? Assuming it's 1 buffer then your total latency is 2048, which is really large for modern hardware. I'm also assuming you're running at 44.1 or 48 kHz (maybe you said...).

I bet if you tested down to buffers where you start getting dropouts that you'd see different results. What you showed is that except for Cubase w/o VE Pro, all the DAWs can run the project just fine at 1024 buffer.

But which ones can run at 96? Which at 128? Etc. I think that's a more useful measurement.

rgames

 

[Nick Batzdorf]

The most important takeaway is that every one of them was able to play the sequence without hurting anyone.

 

[A.G]

Dewdman42 tests are based on a Mac OS X system - thanks for your time D42!

The same test could be totally different on a PC Windows system. In any case it is a hardware specific (partly audio interface, driver etc).

 

[VinRice]

Write in Cubase, mix in Logic. Works for me.

 

[Dewdman42]

Interesting. But 1024 is a pretty large buffer - what VI Pro buffer did you have? Assuming it's 1 buffer then your total latency is 2048, which is really large for modern hardware. I'm also assuming you're running at 44.1 or 48 kHz (maybe you said...).

I bet if you tested down to buffers where you start getting dropouts that you'd see different results. What you showed is that except for Cubase w/o VE Pro, all the DAWs can run the project just fine at 1024 buffer.

But which ones can run at 96? Which at 128? Etc. I think that's a more useful measurement.

rgames

The point of the test was not to see what my macpro was capable of in low latency, that would be another interesting test. The point was to compare the daws to each other with and without vep. 1024 is a good size to use for mixdown of many tracks. You don’t need low latency for that. Anyway vep tends to impose more latency during playback UNLESS you have a track in live mode. Cubase couldn’t even play it at 1024 much less a smaller buffer.

I encourage you to run your own tests at low latency and please share the results.

Testing for ultra low latency brings many other factors into play about which soundcard you’re using and the hardware you are running on, while somewhat fun to find out what your machine is capable of it doesn’t really give us any information that would be useful to everyone. Everyone’s specific setup would be different. I suspect however that the relative performance of each daw compared to the others would probably be about the same as above, but with some hardware you might be able to go lower latency then others with modest cpu, but the comparison between daws would probably be very similar as above.

 

[Dewdman42]

Write in Cubase, mix in Logic. Works for me.

Or maybe mix in protools. But it’s not as much to write for orch if you can’t hear all the tracks playing back. I think vep is absolutely essential for using cubase on a Mac with large track counts (more then 50)

 

[Nick Batzdorf]

But it’s hard to write for orch if you can’t hear all the tracks playing back

We're so spoiled!

There was a time when it was considered wimpy even to use pencil on a score rather than pen, and forget about playing the music on the piano first to hear it. People scoffed at Vangelis - a professor of mine called him Vaginus - for recording the parts to Chariots of Fire directly onto tape one track at a time. Anyone could do *that*!

On one hand, I don't miss staying up all night to copy parts for a session; on the other hand, there's nothing like hearing your music the first time played by real musicians.

 

[VinRice]

I've just finished a 5 minute cue in Cubase 10 with 53 active tracks and 500 additional unloaded kontakt instances. It played back fine and saved very quickly. It would not however Quit after a save, nor would it do a multitrack bounce. I simply did a bounce-in-place for all the tracks and dragged the files into Logic for mixing. I wouldn't dream of mixing in Cubase, there's no way it would stand up.

I found the medium (default) ASIO guard setting is the best - both low and high settings (strangely) would bring it to its knees. I've noticed that opening 3rd Party plug-in user interfaces or indeed Kontakt can start the CPU spike death spiral on occasion. I'll force quit and re-open on a fairly regular basis and that seems to keep thing tickety-boo.

Obviously this is not an optimal situation but it's workable for the time being and I'm not on a particularly powerful machine (2014 4GHz 32gig Quad iMac running High Sierra). I'm sure things will improve with a machine update. I love the writing workflow in Cubase and the mixing efficiency of Logic.

 

[rgames]

The point was to compare the daws to each other with and without vep.

Yes, but the results shows that there's no difference. They all work. Running an orchestral mockup at 1024 buffer is not an issue these days (as you've shown). So, where are the differences? I think they're in low latency performance, particularly on Macs.

It's like testing to see if watching Game of Thrones causes cancer. I presume the test would show that watching Game of Thrones makes no difference on whether people get cancer.

Therefore the more useful question is: what causes cancer?

rgames

 

[Dewdman42]

except Cubase performance was extremely poor with VEP and incapable of even completing the test without VEP at one extreme....and LogicPro performed 10% better without VEP and better then everything else; at the other extreme.

Otherwise, i agree....they were mostly all clustered in the middle with similar performance with and without VEP. StudioOne was slightly better then LogicPro which was slightly better the DP when used together with VEP, but not enough to be concerned about. Cubase performance was horrible at best and LPX alone was exemplary.

low latency performance is a hardware concern.

 

[Dewdman42]

and by all means, if you would like to run tests across different DAW's at lower buffer settings I would be interested in seeing the results, but I have done enough testing for this year...

 

[Geoff Grace]

We're so spoiled!

There was a time when it was considered wimpy even to use pencil on a score rather than pen, and forget about playing the music on the piano first to hear it. People scoffed at Vangelis - a professor of mine called him Vaginus - for recording the parts to Chariots of Fire directly onto tape one track at a time. Anyone could do *that*!

On one hand, I don't miss staying up all night to copy parts for a session; on the other hand, there's nothing like hearing your music the first time played by real musicians.

I remember those days well, Nick. Don't forget about transposing parts for each instrument in the proper clefs, and keeping the concert pitches straight while reading and writing the individual parts and conductor's scores. After centuries of that time-honored approach, we were probably the last generation to do it that way. (Well, maybe a few Gen Xers too.)

How quickly we go from being youthful agents of change to the last of a dying breed...

*Sigh*

What were we talking about again?

Best,

Geoff

 

[rlw]

I’ve just decided to purchase Cubase 2 days ago. My frustration with having to work around live mode in Logic breaks my workflow so badly. I do large orchestra, 166 tracks including aux with VEP. I use spitfire orch with the extended mics and use about 115 gigs of my memory plus 58gig on a PC slave. All my expression, modulation, vibrato etc are done after I lay down my tracks. Sometimes the performance gets so bad that I must draw in my parts. I have a 10 core iMac pro with 128 gigs of ram and a PC 6700 slave with 64 gigs. My templates have one track to a separate Vep session with 4 to 5 kontakt instances and I use Artz Switching for articulations. I have to select a track to get midi window focused on part. Then decouple midi window, then select a muted audio track. So that I can play and edit the CCs properly. My work flow is horrible. I could help my Cpu if I didn't use the extended mics with the Spitfire Orchestra but I love the sound I can architect with the 8 mics. But to have all the articulations available with 8 mics available not only eats up memory but Cpu when I have extensive CC automation. If logic only had a switch to cut off live mode I would be able to make my workflow work. I am very happy with the sound but workflow is horrible for me. Trying to see if my work will improve with Cubase.

 

[JohnG]

I use spitfire orch with the extended mics

That must sound awesome.

Maybe a couple more PC satellite computers? I'm down to three, but it helps a lot. Raises complexity in other ways of course.

Mixing on, and monitoring through, a separate PT Mac also helps since the DAW computer isn't suddenly taxed with mixing and recording tasks on top of everything else.

 

[rlw]

Question.. If I redesigned my template so as to reduce the number of VEP sessions by increasing the number if ports in VEP, does anyone know if I would reduce my Cpu pressure? For instance, I was considering putting all my Strgs in one session by increasing the ports in VEP to house maybe 40 plus kontakt sessions. Would that increase performance

 

[A.G]

Cubase needs a good graphic acceleration and a proper graphic card driver, otherwise it may take more resources for graphics than audio mixing, arrangements etc. Most of the Mac applications need a good QE (Quartz Extreme) acceleration (this is called OpenGL in PC machines). One of the major reasons that Cubase does not runs well on Mac is the GPU.

There are very good PC machines with powerful multi-core GPU which can run Cubase with tons of resources. For example, I own two MSI gaming laptops 17'', which came with very good NVIDIA cards which can be replaced like in the desktops. Cubase Pro ran perfectly with the factory GPUs. Later I replaced the laptop cards with NVIDIA Quadro (CUDA) which totally improved the Cubase performance.

It is known that a few PC Pro Video apps offer an extra CUDA rendering option. If your machine is equipped with a good CUDA GPU then the Video rendering is many times faster than the default CPU rendering option.

 

[rlw]

Cubase needs a good graphic acceleration and a proper graphic card driver, otherwise it may take more resources for graphics than audio mixing, arrangements etc. Most of the Mac applications need a good QE (Quartz Extreme) acceleration (this is called OpenGL in PC machines). One of the major reasons that Cubase does not runs well on Mac is the GPU.

There are very good PC machines with powerful multi-core GPU which can run Cubase with tons of resources. For example, I own two MSI gamer laptops 17'', which came with very good NVIDIA cards which can be replaced like in the desktops. Cubase Pro ran perfectly with the factory GPUs. Later I replaced the laptop cards with NVIDIA Quadro (CUDA) which totally improved the Cubase performance.

It is known that a few PC Pro Video apps offer an extra CUDA rendering option. If your machine is equipped with a good CUDA GPU then the Video rendering is many times faster than the default CPU rendering option.

I have an iMac Pro which has the Radeon Pro Vega 56 with 8GB of HBM2 memory which is a better GPU options on the other Macs. Hopefully this is much better than the other mac computers. However if I go to a PC option that will be something I consider for sure. My biggest concern is how I have configured my templates with VEP.

 

[Dewdman42]

People with newer macs keep telling me that Cubase is running great for them. I wish one of them would do a comparative test as I have done in order to find out for sure. I have a metal GPU and everything else on my mac increased performance substantially under Mojave...but Cubase10 just runs like a dog...and increasingly so with more tracks added, which means...its probably the DSP that is causing the problem more then anything. Move the DSP to VEP and it runs considerably better, though still not nearly as well as the other DAW's with VEP. We can only speculate what the problem with Cubase is, Steinberg has to be the one to figure that out.

If you want to use Cubase on a mac, my recommendation is absolutely use VEP. It can still be used for workflow reasons, just realize its going to use more CPU then the other offerings. But as long as you can play all the tracks you want to play and play with low enough latency..does it matter if the CPU is pegged to 60% vs 40%? Not really. I could not playback 90 tracks on my system without VEP, but with VEP it could play it..though using a lot more CPU..I don't know when and where it would max out... but it could still be perfectly usable if Cubase workflow is what you want and need! But I highly recommend you plan on using VEP with it.