Menu
What's new
By:
Filters

DAW Performance Test Results

Dewdman42 said:
I meant new since vep6
Click to expand...
Ahh ok
I have tried it a few times, but at first I had issues instantiating it, as it would have the red exclamation mark over the plugin all the time

Finally managed to get it working and tested with one instance on one port. Not tried an ensemble yet though, or proper test.

I do like how the port and channel are both in the Inspector like that :)

The results so far are very promising though, so I might find myself back on macOS in the future long term. Sadly one of my reasons for moving to Windows was my CalDigit FASTA6U3Pro is not recognised properly on macOS HS+ when using my Waves DigiGrid D. So it would not let me use both NICs. Windows, of course does not suffer from this
 
Did a new test, this time using LPX with one VEP instance per instrument track.

Drum Roll....

It performs similar performance as other LPX tests, but of the three main LPX tests I did, its the worst, but only by 1% difference, so really this should only be a workflow decision. Here's some data and a new chart comparing three modes of usage with LPX..

pubchart


  1. LPX+VEP+instance-per-instrument: 25% average cpu
  2. LPX alone: 25% average cpu
  3. LPX+VEP-AU3 (single instance): 24% average cpu

As a side note, it took me 2-3x as much time to setup this test, I found the single VEP instance per track to be entirely a PITA in the setup and very laborious to do so. At one point I had to manually reconnect 90 VEP plugins to their various instances on top of it all. Anyway, workflow is different for everyone, the point of this test is to compare performance I would say it does not really make much difference which approach in terms of performance. Use the approach that works best for your workflow.
 
What instrument were you using? And were they VSL? Kontakt?

Dewdman42 said:
Did a new test, this time using LPX with one VEP instance per instrument track.

Drum Roll....

It performs similar performance as other LPX tests, but of the three main LPX tests I did, its the worst, but only by 1% difference, so really this should only be a workflow decision. Here's some data and a new chart comparing three modes of usage with LPX..

pubchart


  1. LPX+VEP+instance-per-instrument: 25% average cpu
  2. LPX alone: 25% average cpu
  3. LPX+VEP-AU3 (single instance): 24% average cpu

As a side note, it took me 2-3x as much time to setup this test, I found the single VEP instance per track to be entirely a PITA in the setup and very laborious to do so. At one point I had to manually reconnect 90 VEP plugins to their various instances on top of it all. Anyway, workflow is different for everyone, the point of this test is to compare performance I would say it does not really make much difference which approach in terms of performance. Use the approach that works best for your workflow.
Click to expand...
 
Dewdman42 said:
Just ran a test using the new AU3 version of the VEP plugin. WOW! Performance beat every other scenario. See new graph on first post of this thread. LogicPro+VEP-AU3 is the new high performance scenario.. and by a lot.
Click to expand...

Did you try Digital Performer with the VEP-AU3? I'm using DP and am curious.

I'm also pretty surprised at your results for DP; for many years, DP smoked Logic in the CPU department because it distributed the load better. Maybe not anymore?
 
I do not know if DP supports AU3, I do not think it does. Anyway I suspect it would be better to use the MAS plugin there.

DP has not ever smoked LPX's cpu performance. LPX has long been known as having excellent CPU performance.

A few years back MOTU came out with their pregen stuff in order to try to catch up their performance with others and tests from other people showed that it basically just brought it in more close to LPX's performance. But VEP can't use PreGen mode, so its a moot point in this case, though I'm not sure right now if my DP test without VEP was using pregen or not.

At the time DP also had very high latency and it turns out MOTU had been using a double sized buffer, which they removed around the same time they added the pregen stuff...which also brought their latency down to normal levels competitive with LPX and other DAW's.
 
But I want to stress, except for Cubase, they are all very close. StudioOne, DP and Reaper are within a few percent. LPX is considerably better though, that is true.
 
Ah, it's all VSL instruments. I wonder if that might be behind the striking similarity in cpu usage?

I'm all Kontakt all the time and from my experience the results would be markedly different with kontakt instruments. VEP saves a tremendous amount of cpu with Kontakt and many others.



Dewdman42 said:
Please read first post on the thread.
Click to expand...
 
Well its all relative. It would be great to have another test project using Kontakt or something everyone could try, but I don't have time to do that, I'm just about burned out on all this performance testing.

The test I did played exactly the same section of music through the same instruments and FX in each situation. That's why the graphs you see all follow the same curve, but somehow, certain DAW's or configurations need a little more CPU overhead to do what they do. The point was to compare DAW's not plugins. If we used kontakt instead of VSL instruments, I would expect to see some kind of CPU graph over the course of the same music...and each DAW scenario would follow a similar curve, with probably similar differences between each scenario. Its hard to say whether Kontakt would overall use more Cpu then ViPro, but that isn't the point of the test.

Some more interesting tests would be to just make something that cranks out an endless stream of 8th notes or something and then see how many tracks you can get before dropouts with each DAW combination. you could do that for both VSL and for Kontakt...and I would expect that DAW shootout results to be about the same, but you might find out that kontakt is more or less efficient then ViPro..but its a moot point. If you're using kontakt libraries you don't have a choice and same with VSL. What we do have a choice about is which DAW to use and whether to use VEP or not. Which is more the point of my testing. I did not even really push my rig at all even close to the limit, it could easily handle a lot more tracks then I tested I'm sure. well, not under cubase. But under the others it could. And other people's rigs could handle even more. That's a different test too. This test was really about comparing the DAW's to each other with a real world orchestral mock up...and comparing them with and without VEP involved.
 
Dewdman42 said:
DP has not ever smoked LPX's cpu performance. LPX has long been known as having excellent CPU performance.
Click to expand...

Maybe I was not specific -- the only area in which it outperformed logic was distributing processing among multiple cores. I wasn't saying that "per CPU" it was better.

But that was true for years. VEP kind of made it a moot point.
 
charlieclouser said:
Also I think Logic is able to do some sort of pre-processing, almost like an invisible-to-the-user freeze function, so that tracks / instruments that are not record-enabled are being calculated when the transport is stopped, so that when playback begins the cpu is just spooling a pre-rendered file instead of calculating the instrument + plugins in real time.

Not sure if that is still a part of Logic's audio architecture, but for sure the "dual buffer" thing is, where Logic calculates all non-record-enabled tracks at some obscenely huge buffer while all record-enabled tracks are calculated at the actual buffer size you've set in Preferences - and stuffed onto that last cpu core as well. So, some pros and some cons with that approach.

But I am always amazed at how efficient Logic is. My biggest sessions that use only EXS and audio can have hundreds of instruments hammering away at 16th notes and they play equally well on my 12-core cylinder and on my 2012 quad-core laptop, rarely going above 50% cpu.
Click to expand...

Efficiency is the main reason I switched from Cubase to Logic. And now Logic 10.4.5 is absolutely fantastic.
 
Alright people, the plot thickens. See new chart in the first post. Cubase 10.0.30 is performing significantly better on the mac than the previous version. The previous version without VEP could not even play this project without crapping out audio 20 seconds into it, and even with VEP7 involved, it was just barely playing the project with average cpu usage at 49%, much higher then all other DAW's tested. The 10.0.30 update has dropped average cpu usage to 33%, which is better than several other DAW's, but still not better than LogicPro. Without VEP its only 1% worse at 34%, with ASIOGuard turned off, and 35% with ASIO Guard turned on (normal). All well within the range of usability.

See OriginalPost for more details

pubchart


pubchart


Bravo Steinberg!
 
Last edited:
Thanks for all of this work, it’s illuminating.

My only problem is with the 1024 buffer. I play everything in live so that’s just not an option for me.
 
You must log in or register to reply here.
Top Bottom