Dynamic Plugin Processing Not Working on Audio Tracks - Killing My CPU

[tmuzik]

Is this an "Only Me" thing?

I am particularly sensitive to CPU usage b/c I have one of those 2017 i7 iMacs that gets hot quickly and the fan noise is horrendous.

This issue is with audio tracks - NOT Aux returns.

Out of many thousands of dollars worth of plugins only a few brands are now working correctly. AVID's own plugins do not work. Let me preface in saying I did not test them all - so far Pro Comp, Pro Limit, EQ7, Revibe II, D-Verb.

I hope I have a tick-box some where unchecked, and I am just being dumb - but being that 3 brands are working on audio tracks and the others are not would lead me to believe that it is not a settings issue.

ALSO the same plugins that DON'T work on AUDIO tracks DO IN FACT work on AUX returns - again leads me to believe it's a Pro Tools thing.

I have simple session called "128 track CPU test" that I use for seeing how resource intensive a plugin is before I purchase.

SEE VIDEO

https://youtu.be/3RL-709IIwM

THE SESSION SETUP

• 128 stereo tracks - in 2 groups of 64 - no input - output to main 1-2
• the stereo tracks all have audio for 1 measure --- 30 second gap of nothing --- than audio for another 4 measures (this is whereI can see how long it takes a plugin to become inactive) - some are actually very fast - brands range from 2-15 seconds
• 100 aux returns - all have the same "shared" aux input - (so I can feed all of them with one send) - output to main 1-2
• 24bit - 44.1 - 128 samples - dynamic plugin processing checked -

TO TEST AN AUDIO TRACK

• first pass - using no plugin's as a CPU baseline - audio only -
• second pass - I load plugins only on the first 64 audio channels and monitor the system usage -
• I then add or subtract plugins to find out the maximum my system can handle -

TO TEST AUX REURNS

• first pass - using no plugin's as a CPU baseline - audio only -
• second pass - I load plugins across the aux channels and monitor the system usage -
• same setup as before - except this time I send from an audio track into the shared aux
• I then add or subtract plugins to find out the maximum my system can handle -

Such a simple to make - handy template - Last time I paid attention to the "dynamic processing" aspect during test's was probably 3rd quarter 2020.

Kind of ironic - Logic has the exact opposite problem (last I checked) - Audio tracks, dynamic plugin processing works fine - but the aux return plugins are always active -

Anyway - this is kind of a serious issue - I may have many, many tracks with only a snippet of audio. It is sad to think of all those tracks with a tiny snippet are making my processor work so much harder.
If I am looping one section that may only be 10 audio tracks, EVERY PLUGIN PROCESSOR FIRES UP ACROSS THE ENTIRE SESSION and Does not stop until until the playhead is at rest. (except for those brands mentioned below)
This is like 2004 all over again --- I guess I need to sell my car, and my wife's, then borrow some cash from some questionable friends and get an HDX rig?

THE BRANDS THAT WORK CORRECTLY
The only plugins brands that seem to work on my system are Slate, Sound Device Digital - Melda - with the time to sleep being 15secs for Slate, 4sec for Melda, 2sec for Sound Device Digital (good job Sound Device Digital).
I didn't test every brand... I don't have that kind of time.

THE BRANDS THAT NEVER SLEEP
AVID, Plugin Alliance, Waves, TDR, Karzog, iZotope, Tone Empire, Arturia, "a WHOLE LOT of Goto" plugins do not work dynamically on my system.

THINGS I TRIED

• disabled HEAT
• Removed HEAT plugin from the systems plugins folder
• Optimize perf for low buffer on/off
• Ignore playback errors on/off

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

Pro tools 2020.12
Catalina 10.15.7
iMac
Model Identifier: iMac18,3
Processor Name: Quad-Core Intel Core i7
Processor Speed: 4.2 GHz
Memory: 64 GB
Boot ROM Version: 428.0.0.0.0
SMC Version (system): 2.41f2

[albee1952]

Tough to say what's going on, but I suspect its fairly unique to your setup(way too many folks having solid performance to blame this on PT). Does your iMac have a Fusion drive? Those can be a performance killer. What drive is hosting your sessions(and exactly how is it connected)? As for Dynamic Plugin Processing, its never been a game changer of any sort(I'm still unconvinced that its worth anything). Speaking from my own experience, I rarely have any issues with plugin performance other than how long it takes for Melodyne or Ozone to open up on the screen(but our systems are quite different)

You might also check to see if all your plugins are the latest versions

[tmuzik]

Originally Posted by albee1952

Tough to say what's going on, but I suspect its fairly unique to your setup(way too many folks having solid performance to blame this on PT). Does your iMac have a Fusion drive? Those can be a performance killer. What drive is hosting your sessions(and exactly how is it connected)? As for Dynamic Plugin Processing, its never been a game changer of any sort(I'm still unconvinced that its worth anything). Speaking from my own experience, I rarely have any issues with plugin performance other than how long it takes for Melodyne or Ozone to open up on the screen(but our systems are quite different)

You might also check to see if all your plugins are the latest versions

Yes all plugs are update - all paid for -
I can't think of a reason hard drive speed would have anything to do with dynamic processing - No Fusion drive here - SSD Internal - super speedy.

This was working a year ago - in fact I keep a document which has all my test data and the test I ran a while back only had one brand not respecting the dynamic plugins. Now it's like all of them.

Dynamic Plugin Processing is a HUGE deal!

If I have 64 channel session - as you can imagine there are not 64 tracks ever playing simultaneously - However without Dynamic Plugin Processing they all might as well be playing at the same time, b/c the plugin will use the processor even if the track only has a single 1 second piece of audio on it. So if you are looping a section that has 9 tracks - and working in a new keyboard part - then EVERY AUDIO TRACK in the session will fire up their plugins

@128 sample buffer
PA - Lindell 50 x 64 channels - no oversample - use 20-22% on my processor
Kazrog - True Iron x 64 - zero latency - no over sample - 48-50%
Now add some reverbs, delays and saturators, master bus tools.

It grows quickly - especially if you consider the biggest CPU hit is in the first instance of many plugins in the effects category -
Waves Abbey Chamber takes my processor from 6% to 28% for the first instance -
The second instance bumps up to 35% - so 10%-ish for each of the following instances -
However I only get the CPU penalty when audio is running through it + 10 seconds. Because the returns work as expected.

So if I used the Chamber just for the guitar solo in the middle then no big deal - nothing else is going on - take the processor hit for the center of the song.

However - if a track with an Abbey Plate is following the solo then I might run into issues - so I might opt to use two Plates or two chambers to save the 12% CPU hit of a first instance.
Or freeze it - but having to freeze tracks - is a last resort for me.

The number seem high above but that is because the buffer is still kind of low - at 128 samples. Another topic for another day - a bigger buffer than 1024 would be so welcomed for mixing.

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

None the less the plugins DO WORK AS EXPECTED on Aux channels - but fail miserably on audio tracks. So one could deduce that the plugins are not to blame here.

[bolooki]

does automating bypass make any difference here?

I personally gave up on this option because everyone said it didn't do anything. good do know it works on some plugins and it may be worth asking those companies tech support why that might be.

[tmuzik]

Yes I don't know why so many say it doesn't do anything. It was working months ago for me and is not working now. To say it didn't work is just plain ignorant to facts. The video posted below shows this feature working 5 years ago. There are also many mentions on the pro-tools-expert.com website of the feature and how to use it, and being that "expert" is in their domain name clearly means they were early adopters of this intra-webby-net thingy...

I had no issue in the past with Waves, Avid, Karzog, Overloud, Klanghelm, IK plugins.

Plugin Alliance, Slate, TDR, the United Audio guys are new to me so I never tested them for CPU hit. I did make note during my last round of testing that Arturia Pre bundle was "DRDP" which is an acronym I made for "Doesn't Respect Dynamic Plugin"

Now it's like almost all brands crap out, but only on the audio track. They all work PERFECTLY on aux's... weird...??? (head scratch)

Something must have changed in the code - maybe - BUT nobody else has chimed in and said they see the same issue.
SO maybe I am getting old and senile - that's a fact too.

Here is a video from 5 years ago explaining dynamic processing and you can clearly see it does work.
https://youtu.be/v9ISARiE2oQ

I did learn from the video above, if a plugin is not going into "sleep mode" (for lack of a better term) - like the analog emulation type that have constant background noise and hiss... which I still find extremely funny... that the noise plugin prevents the entire chain even if it's the last plugin in the chain which theoretically should not effect the plugins that precede it but it does... nice to know.

I find it critical in my plugin purchases to know how much mileage I can expect from any one plugin. This all started years ago with my max out the CPU stress test. How many of instances of any one plugin can I use. Which then gives a solid baseline for plugin choice, depending on the complexity of the session. Which plugins I should reach for if I want to be CPU conservative and which plugins are going to be fun but gobble my processing.

Maybe I will take this question to Fabien from TDR he is fairly active on the GS forum...

[Cheesehead]

Interesting thread!
You would think that having dynamic plugin processing on would decrease cpu load.
But I became suspect of it too and since I've turned it off in all my sessions I have a lot less crashes and spinning beachballs.
I think if you have a powerful enough machine it might be actually detrimental to have it switched on.
Maybe if it worked correctly that would be different?

[tmuzik]

Well there seems to be no doubt that when it works it does decrease CPU load.

I remember when I was running a Native Rig on a Windows Xeon box - I would get weird CPU spikes, that were so difficult to track down, I would blame anything and everything and lose sleep at night until it was sorted. (Dumb windows 10 forced update.. GRRRR....)
I think some people have issues and guess what they think the problem is, or change 20 things, which fixes the problem and can only speculated on what the actual fix was... I know I have done that. Especially with BIOS changes.

I would be really interested if someone had a method of showing the phenomena of dynamic processing being a problem, in a way that can be re created, further tested and refined.

Yes please if some one has insight on why audio tracks mysteriously stop being part of the dynamic plugin processing for most plugins - I would be very grateful.... and maybe buy you a beer, or a yoohoo.

Does anybody personally know a plugin developer?

Is it broke on Windows?
Pro Tools 2018?
Pro Tools 2012?

At the very least with the current version of PT the aux's do work beautifully.
So knowing that big CPU reverb i.e. Waves Abbey Plate or Chamber won't be processing the entire length of the track is nice.

Any cracked plugs that I had pirated are trashed - Huge stability gain (as you can imagine) being a pirate is a lame feeling and is thievery.

One last thing to note - In my latest testing - I went through some lesser known plugin developers from freebies I received at Plugin Boutique - and one of them was extremely impressive in releasing the CPU in 2 seconds - which is sick-O improvement over the Slate VMR that takes it's good old time of 10-12 seconds. Not a dis on Slate at least his plugin respects the Dynamic management on audio tracks, but 10 seconds seems a bit long.

But I KNOW Waves did dynamically process at one time and now they don't - so maybe it's something the plugin manufactures did for the Big Sur compliance mess or maybe something AVID changed or both.
I don't know?


Is there a performance hit with the plugins turning on and off? If there is I don't see it. I don't see a cpu spike at all on any tool I use to read my CPU, even when 512 plugins turn on than off than on.

If someone has a better way to test or see's I something I didn't consider, please do let me know.

I don't have any problem with changing my view point to better reflect reality.

[JFreak]

... and here comes the most important question of them all, what PT/OS combination you are using and are your plugins supported on that combo?

[Darryl Ramm]

And when somebody states they have lots of plugins and are having *any* crazy problems the first question should be have you removed .aaxplugin files from the system systemically looking for badly behaving plugins? That might not be the causes but it has to be a leading suspect so you have to troubleshoot. Even if everything is up to date and supposedly supported (but that should be known as well). I can't even make it though all the posts here, but I don't see plugin removal/testing leaping out here.

I'm not even sure why this is talking about dynamic plugin processing. If you are throwing empty audio into random plugins then uh not surprisingly you might be getting some denormalization problems trashing CPUs. All depends on the plugins.

Dynamic plugin processing can interact with different plugins differently anyhow, in some cases it can cause instability. Again, troubleshoot the plugins, (and look for denormalization issues).

[tmuzik]

Vanilla Pro Tools Test:
https://youtu.be/izsM36eYyDM
Video description has timecode labeled contents - so you can speed through it.

This test consisted of Pro Tools Stock plugins only -

• Avid Channel Strip - dynamic processing works on aux's
• Avid Channel Strip - dynamic processing doesn't work on audio channels
• AVID - D-Verb - Same as Channel Strip
• AVID - EQ-7 - Same as Channel Strip

Until a developer that understands the AAX codebase chimes in - I can only speculate. I can't dumb down the test anymore. All software variables were Avid.

If it's working for everybody else on 2020.12 mac version or there is a tick box to disengage the dynamic plugin feature on audio tracks, please let me know what I am missing here.

I may take a peek at the AAX SDK - to see if I can find the proper name for the variable that triggers the dynamic processing. If it's like any other script you would need to call a variable and that variable would have to be defined so the developers know how to initialize that feature or function.

included the specs... again ;-)
------------------------------

Pro tools 2020.12
Catalina 10.15.7
iMac
Model Identifier: iMac18,3
Processor Name: Quad-Core Intel Core i7
Processor Speed: 4.2 GHz
Memory: 64 GB
Boot ROM Version: 428.0.0.0.0
SMC Version (system): 2.41f2