AAE-9173 - Ran Out Of CPU Power

[Guilla]

Thank for bringing me to this thread man !

It is crazy it seems than you are working harder than Avid to find a solution to this problem !
Avid should be ashamed...

Anyway it happens to me also, at such a point that my session is unreadable (spike every 5 seconds). I have done every thing.
Sometimes lowering the mixing and editing window helps, but not in the session I got currently.

Setup :

PT 12.6.1
Macbook Pro mid 2015, 8 month old Sierra 12.10.1
16 Go RAM, 512 Go SSD
Mixing from a 1 To HDD 7200 or 500 SSD (doesn't make a difference).

My CPU monitor activity told me that 49 % of the CPU is free when PT CPU spikes like crazy.
It become always worse when I create my Mix Bus and print track.

I'll run your test tonight if I find the time and post the results.

It is driving me crazy I have got a session for a client that is clearly unreadable. I'll have to freeze every tracks to try to read it.
Remember me when I started mixing on a really ****ty computer I bought for 200 bucks on a ****ty DAW...

[Burbree]

Thanks Wonderboy,

It would appear that Auxiliary workflows are definitely contributing to what are causing problems.
My system would play over 350 Pitch II's on Audio tracks but once it hit 80 instances on Auxs.. 9173s

Just as a sidebar to prove that not all plugins are written the same: Look at this screenshot:

Untitled.jpg

Yep thats only 4 BX Meters running on your session on Auxs and its already falling over..

[dankin]

Originally Posted by Burbree

Thanks Wonderboy,

It would appear that Auxiliary workflows are definitely contributing to what are causing problems.
My system would play over 350 Pitch II's on Audio tracks but once it hit 80 instances on Auxs.. 9173s

Just as a sidebar to prove that not all plugins are written the same: Look at this screenshot:

Attachment 9277

Yep thats only 4 BX Meters running on your session on Auxs and its already falling over..

I'm going to do some experiments this week. It definitely appears the aux's are a big part of the problem. So do all aux's operate on the input buffer even when fed by a internal buss? It would make sense they would use the input buffer when fed by a hardware input, but not when being fed an internal buss. Although with that being said -if they are running on the input buffer at all times, why does the input buffer seem to have no affect on the 9173's? I tend to leave mine at 512, unless I'm recording a VI or monitoring audio through PT. But, at 1024 I still get the same spikes.

My vocal channels and mix buss tend to have the most CPU heavy plugins. For the mix buss, I use 2 auxs. The first one has my 2 buss processing, the second one has my limiter. This is partly so I can print 2 versions at the same time, one with a limiter for refs, and one without for the master, but it also spreads the load CPU load out.

[Guilla]

Originally Posted by dankin

I'm going to do some experiments this week. It definitely appears the aux's are a big part of the problem. So do all aux's operate on the input buffer even when fed by a internal buss? It would make sense they would use the input buffer when fed by a hardware input, but not when being fed an internal buss. Although with that being said -if they are running on the input buffer at all times, why does the input buffer seem to have no affect on the 9173's? I tend to leave mine at 512, unless I'm recording a VI or monitoring audio through PT. But, at 1024 I still get the same spikes.

My vocal channels and mix buss tend to have the most CPU heavy plugins. For the mix buss, I use to auxs. The first one has my 2 buss processing, the second one has my limiter. This is partly so I can print 2 versions at the same time, one with a limiter for refs, and one without for the master, but it also spreads the load CPU load out.

That's freaky crazy, only 4 instances !

[wonder6oy]

Originally Posted by Burbree

Thanks Wonderboy,

It would appear that Auxiliary workflows are definitely contributing to what are causing problems.
My system would play over 350 Pitch II's on Audio tracks but once it hit 80 instances on Auxs.. 9173s

Just as a sidebar to prove that not all plugins are written the same: Look at this screenshot:

Attachment 9277

Yep thats only 4 BX Meters running on your session on Auxs and its already falling over..

Wow - and I thought my mere 8 instances of Scheps PP was rough. It appears from your screenshot that you are using a native version of Pro Tools, is this correct?

[Guilla]

Originally Posted by wonder6oy

Wow - and I thought my mere 8 instances of Scheps PP was rough. It appears from your screenshot that you are using a native version of Pro Tools, is this correct?

I am using Native Pro Tools too btw.

[wonder6oy]

Originally Posted by dankin

I'm going to do some experiments this week. It definitely appears the aux's are a big part of the problem. So do all aux's operate on the input buffer even when fed by a internal buss? It would make sense they would use the input buffer when fed by a hardware input, but not when being fed an internal buss. Although with that being said -if they are running on the input buffer at all times, why does the input buffer seem to have no affect on the 9173's? I tend to leave mine at 512, unless I'm recording a VI or monitoring audio through PT. But, at 1024 I still get the same spikes.

My vocal channels and mix buss tend to have the most CPU heavy plugins. For the mix buss, I use 2 auxs. The first one has my 2 buss processing, the second one has my limiter. This is partly so I can print 2 versions at the same time, one with a limiter for refs, and one without for the master, but it also spreads the load CPU load out.

I'm not really sure about that notion that Aux's are handled in the same manner as live inputs (it doesn't really make sense to me either); I personally think the problem is something else, but only Avid would know. I have been able to confirm however that the longer the signal path is (Aux to Aux etc. to Audo track to Master) coupled with processing along that path, the more likely the spiking behavior is to occur (and the lower the CPU usage number before spiking is likely to be). So splitting out mix buss and master buss processing on separate Aux's might be exacerbating the issue further. Just a heads up.

[wonder6oy]

Originally Posted by Guilla

I am using Native Pro Tools too btw.

Thx Guilla! Out of all who have come back with results so far, (I think there are 5 now), the only one without this issue was an HDX system. The rest are Native systems (though I am unsure if there are any Native HD system results yet).

Don't know if that's enough to call it a "Native-only" issue yet, but... possibly?

I suppose I should ask for that info in any further results.

[treason]

Originally Posted by wonder6oy

Thx Guilla! Out of all who have come back with results so far, (I think there are 5 now), the only one without this issue was an HDX system. The rest are Native systems (though I am unsure if there are any Native HD system results yet).

Don't know if that's enough to call it a "Native-only" issue yet, but... possibly?

I suppose I should ask for that info in any further results.

Speaking for HDX users - my normal path consists of cascading a signal through approximately 10 busses serially, and that same signal is also split parallel at the source (and farther along) another 3-6 times.

Correct that HDX uses its DSP card for the mixer (as I noted in previous posts and images, I use approximately half of my 18 HDX DSP processors for handling my typical session routing), and I have 439 active busses in my sessions.

Even if I run my plugs all Native, the mixer still instantiates on DSP and it is my experience that this has allowed me to create complex sessions not possible in a native only version of Pro Tools (not to mention delay compensated side chains).

If I have other applications running (Safari, DropBox client, iMessage, etc) I do notice an increase in the potential for -917x errors, however if I run Pro Tools and some Apple note taking apps (Pages, Numbers, Notes), my CPU usage is almost flat ... that meaning it fluctuates within a single digit percentage (81-88 for example).

[dankin]

Originally Posted by wonder6oy

So splitting out mix buss and master buss processing on separate Aux's might be exacerbating the issue further. Just a heads up.

For me, it seems to help because it spreads the load on the cores better. This is more of a deal if you have several CPU heavy plugins on one channel. If they are all light, it may be better to keep them on one.. But, who really knows? It sucks that we have to constantly trouble shoot this crap.