Horrible VI performance when recording in 2019.10!

[RyanC]

Let us know when you have the test together. I'm looking forward to trying it out.

With the newest policies of Avid, I'm considering my options...

One of the biggest factors for me is performance in the context of a template that includes buses and FX send/return buses and otherwise full routing. Will this be covered in your test sessions?

If not that will still be very useful to me.

[TNM]

Originally Posted by RyanC

Let us know when you have the test together. I'm looking forward to trying it out.

With the newest policies of Avid, I'm considering my options...

One of the biggest factors for me is performance in the context of a template that includes buses and FX send/return buses and otherwise full routing. Will this be covered in your test sessions?

If not that will still be very useful to me.

as far as busses and returns the only test that will utilize this is the "armed audio tracks" test, which is designed to see how the DAW performs seding reverb busses to audio tracks that are armed and monitoring at low latency, whilst the tracks also have insert effects of their own on each of them.

It's too complicated to add it to the VI tests because it makes the basic "how many tracks of this VI can you get" which is an easy way to easily balance load all cores perfectly evenly (will be just a 4 polyphonic chord progression repeating over a 2 minute length and the DAW must be able to play back the entire length without glitches), well, not so basic anymore LOL.

In the audio input monitoring tests it's essential OTOH, as it's a quality reverb that is the hardest thing to run at low buffers. As far as busses to busses and things like that, we'll see how we go down the track.. It will be very streamlined to begin with, but as with anything of this nature, anyone is welcome to alter it for their own needs however they see fit. I'll be following the tried and true DAWbench formula for VI's which is just duplicated tracks playing the same thing and the same patch of the same VI.

I think kontakt player is the way to go for the VI test but I am still open to suggestions on other freeware.. I just need to find a patch that is cpu taxing and is freeware.
I designed the new Logic benchmark test over at GS (same user name), so that will give you some idea of how I construct them. Cheers!

[JFreak]

Did you already check which version of Kontakt or VEpro you are using? Pretty important factor here...

[TNM]

Originally Posted by JFreak

Did you already check which version of Kontakt or VEpro you are using? Pretty important factor here...

Huh? You are on the wrong track.

The performance differences I have picked up have zero to do with kontakt.. Kontakt only came into the discussion just now as a possible freeware (the player) to use in construction of the standardised performance test..
Ve pro? Will never buy or use that useless thing in my lifetime, sorry.. On the same computer it offers much worse performance than just using the DAW, and also adds latency if you want the performance to be at least decent.. It's totally unsuitable for my use and the results or posts here have nothing to do with VE Pro.. they are just projects in general with a variety of synth based VIs that can't play back anymore as PT 2019 doesn't have the cpu overhead that previous versions had. Because I noticed this, I created some benchmark tests using a couple different random plugin synths to confirm it, and indeed I did.

[JFreak]

Just saying v1 VI on v1 DAW may perform differently than v1 VI on v2 DAW. Always check compatibility and use latest version for a given DAW

[cmbourget]

Originally Posted by TNM

Also I want to ask a favour.. I asked this last time and no one would help.. It really won't take much of your time cause you launch pro tools every day anyway!

If you are on mac and using core audio, PLEASE..

set your buffer in PT to at least 64.. go 32 if you are game.

Load ANY VI known to be resource heavy (things like Xpand will not cause issue).

Press the record monitor button. Just tell me if the pro tools meter starts suddenly loading all cores for just the one instrument armed.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

I have some more results with Sylenth synth.. this is actually even more alarming as I did the tests this time with record OFF also.

Sylenth patch "Trance Epic".
3 note, 4 bar chord, copied for 16 bars total and looping. Same midi chord file on all 3 DAWs.

Buffer on all DAWS = 128 this time, to give PT a fighting chance.

because PT's internal playback buffer is 1024, i set studio one to 1024 (high setting) and logic to 1024 (the medium setting).

NON Record armed tracks:

Logic 10.4.6:
71 instances!
Maximum Real CPU used: 565% out of a possible 800%.

Pro Tools 2019.10:
32 instances! WTF?
Maximum real cpu used: 563% of a possible 800%.

Studio one V4.5:
Unlike Logic and Pro tools, studio one never stops playback on an overload.. it just does little distortions that worsen the more the cpu is overloaded. The stable result is 35..That's the number I confirm has absolutely no clicks or pops.
Maximum Real CPU Usage = 329% of a possible 800%.
S1 is therefore *unbelievably* inefficient, as it always has been..
Pops and clicks would happen out of nowhere even on 40 instances, and S1's own usage meter had headroom! I used the AU just like Logic.

When record is armed:

Logic 10.4.6:
23 tracks!
This, despite Logic only using 4 of the 8 threads when tracks are armed (I think this is actually very clever cause it leaves half the threads clear for playback tracks).

Studio One V4.5:
5 tracks! S1 is just SO bad at live VI recording.. Even on a moderately busy project, just arming one VI sends it going nuts.. The unbelievably good track transform helps alleviate it to a degree cause you can one click choose between audio and midi for a track.

PT 2019.10:
19 tracks! it did very well this time!


Now ok, the recording tests for so many VI's may not be real world but playback surely is.. Deciding to included the real cpu usage this time has opened my eyes a bit.. Is Sylenth AAX using a lot more CPU than Sylenth AU/VST? Because PT is able to use the same REAL Cpu as logic before playback drops out. SO.. i'm going to put this to the test by using the AU wrapped in Blue Cat Patchworx.

However, if there is an AAX native plugin you KNOW is uber efficient and programmed well.. please let me know so I can put it to the test!

Ps The reason I used the real cpu usage only on the playback track test was because that's the one I feel was relevant and matters in the real world.. Just how much of your cpu can your DAW really use?
Also of note, I was able to get around 680% real cpu usage in PT 12.8.4 on the same macbook with sierra.

I prefer to work with Pro Tools, but I also have Cubase 10.5 and I'm surprised by its ability to handle latency. Very powerful. You should try in your tests. You would be surprised.

[TNM]

Originally Posted by cmbourget

I prefer to work with Pro Tools, but I also have Cubase 10.5 and I'm surprised by its ability to handle latency. Very powerful. You should try in your tests. You would be surprised.

I have.. Cubase performed very well in the multi track low latency record armed track test. very well.. It goes, Logic followed by Reaper than Cubase, then bitwig.

All other mac DAWs are way behind with S1 being less than half of the top 3.

I think Cubase mac is actually, for the first time ever, running better than the windows version, cause windows has the MMCSS issue that is still plaguing Cubase 10 (when tracks are record armed, cubase can't use more than 6 cores by design now with Cubase 10 on windows, which they call a "fix").
Cubase 10 does work well on windows when using asio guard and for playback tracks though.. but live track performance has stunted sadly.. and since there are no such MMCSS limitations on mac, theoretically it should beat windows now. I'll find out myself on my imac pro soon in bootcamp... 8 cores/16 threads is just high enough to trigger the MMCSS issue which occurs above 14 threads.

That said, why not answer my question you quoted?

All I am asking for is one person to arm one VI track on their MAC (windows is ok) with PT 2019.10 at 32 buffer, and see if all cores start getting used rather than just one.

Anyway, I can't change the headline but this topic's situation has changed somewhat.. PT low latency performance is no worse for me than it was in 2015.. I originally made this topic cause support told me they had fixed that particular issue.
But I have always worked around it by using UAD for external input monitoring, and 128 buffer in PT to play VI's..
The issue is now, that the playback buffer performance has drastically decreased which is far worse cause i means I can get less VI's and plugins than before.

[TNM]

Originally Posted by JFreak

Just saying v1 VI on v1 DAW may perform differently than v1 VI on v2 DAW. Always check compatibility and use latest version for a given DAW

Sure they can..

My tests are always done with the same plugin version..

It is not the VI's fault or the dev's fault if their VI changes performance for the worst in a PT update if it was working before.

It's not even a consideration though, as then I would have to throw out every plugin. I have tested 20 "brands" so far and every one suffers from the new bottleneck..I am not going to test every plugin I own. Fab filter plugins have the best aax coding around and they too are "afflicted".. every VI is.

I have some new video now from someone that HD native does the same thing at it's lowest buffer, even trying to play one VI to record the midi notation, all cores start spiking..
Anyway at least now I know it's not just a core audio issue, and i am guessing the same would happen with native plugins in HDX.
I still don't know where VE Pro fits into this.

If your reply was talking about the initial post relating to low buffer record armed tracks, then why not do the simple test? Put your buffer at 32 and arm a vi track and start playing notes and tell me what happens, if all cores spike for the one VI.
Here's what happens in mine.. As you can see at the top in the istat cpu indicator, those cores really are being used.. I can show you in other DAWs how only one is used when one vi is armed.. there is nothing else in this project...
I can give you pictures of 50 different VI's including the AIR ones (except xpand cause it uses no CPU) and they all do it.

[musicman691]

TNM:
Is there any real, legitimate reason you work at such a low buffer?

[cmbourget]

Originally Posted by TNM

I have.. Cubase performed very well in the multi track low latency record armed track test. very well.. It goes, Logic followed by Reaper than Cubase, then bitwig.

All other mac DAWs are way behind with S1 being less than half of the top 3.

I think Cubase mac is actually, for the first time ever, running better than the windows version, cause windows has the MMCSS issue that is still plaguing Cubase 10 (when tracks are record armed, cubase can't use more than 6 cores by design now with Cubase 10 on windows, which they call a "fix").
Cubase 10 does work well on windows when using asio guard and for playback tracks though.. but live track performance has stunted sadly.. and since there are no such MMCSS limitations on mac, theoretically it should beat windows now. I'll find out myself on my imac pro soon in bootcamp... 8 cores/16 threads is just high enough to trigger the MMCSS issue which occurs above 14 threads.

That said, why not answer my question you quoted?

All I am asking for is one person to arm one VI track on their MAC (windows is ok) with PT 2019.10 at 32 buffer, and see if all cores start getting used rather than just one.

Anyway, I can't change the headline but this topic's situation has changed somewhat.. PT low latency performance is no worse for me than it was in 2015.. I originally made this topic cause support told me they had fixed that particular issue.
But I have always worked around it by using UAD for external input monitoring, and 128 buffer in PT to play VI's..
The issue is now, that the playback buffer performance has drastically decreased which is far worse cause i means I can get less VI's and plugins than before.

I am now on PC, so I can not help you. But thank you for your technical explanation. A little complicated for me, but hey, I follow you.