HDX PT 2018 CPU Spikes AAE -9171 Halts Playback

[Howardk]

Originally Posted by the.engineer

Any major features missing?

And did you get your plugins working?

Unfortunately it looks like Spectrasonics current versions and Yosemite are a no go (link to compatibility info). I use Omnisphere 2.5, Keyscape, Trilian and Stylus. . . which means no PT/OS X version to land on that does not have Pro Tools CPU Spikes, and is a Spectrasonics compatible OS X version. I could install older versions of these VIs, but then I will likely not be able to go back and forth between PT/OS versions as I was hoping. . .

Current version of Native Instruments Komplete 12 is not compatible with Yosemite (Native Access will not work) and the compatible version of Native Access (1.51) is crashing when I try to set the library folder location or link to some libraries. . . I sent a post to NI.

Serum Installer did not install the Serum Presets. . . had to drag those over from another system drive that already had Serum installed, but at least that is now working.

Problems 3 for 3. . . it's a bit of a mess so far.

[LDS]

Originally Posted by Howardk

I am curious if anyone with a 2010 or 2012 Mac Pro can run Pro Tools 2018 with some native plugins and small buffers without similar problems?

I just tried the test session. Opened it, set the buffer to 64, record enabled the record tracks group, recalled memory location 2. My system comfortably recorded the full 118 minutes with the CPU meter sitting around 30% - occasionally oscillating between 22% and 45%.

I've attached a screenshot. Have I done it correctly? To push it further, I decked out the first 5 plugin slots on the first 5 channels with other various native plugins including a bunch of Plugin Alliance plugins. It comfortably recorded to about 10 minutes before I hit the stop button. CPU meter around 52% or so. I have no reason to think it wouldn't continue for the full 118 minutes.

Interesting, the Eleven MKII plugin really messes up the distribution of processing power across the CPU cores. I've always noticed a bit of a peak in the centre of the Usage meter, but it is nuts with the Eleven MKII plugin. The actual total CPU usage is reflecting the highest core, despite there being a whole bunch of spare, unused processing power on the other cores. Replacing replacing Eleven MKII with guitar rig instances seems to immediately reduce the CPU load by about 7 or 8%, and it distributes the load a little more evenly across the cores.

Did I do the test correctly? There seems to be a pretty sizeable difference between my system and yours in terms of performance... even though they are pretty close in specification. James touched on the issue being influenced by video. I am running a bunch of pretty old HP LP2275w monitors with a resolution along the lines of 1600×1200. Perhaps that has a sizeable influence on performance?

EDIT. Turns out those screenshots are tiny. Here are some dropbox links.

https://www.dropbox.com/s/1wkmytip80...%20pm.png?dl=0

https://www.dropbox.com/s/9gw6xrpa49...%20pm.png?dl=0

[Howardk]

Originally Posted by LDS

Did I do the test correctly? There seems to be a pretty sizable difference between my system and yours in terms of performance... even though they are pretty close in specification.

Looks like you did it correctly to me. Yes, big difference in performance. I can't even run a 64 buffer for more than a second or two with that test session. I am mystified.

Have you experienced the other issue I have, extremely slow response to solo playist lanes (to audition takes)?

[LDS]

Originally Posted by Howardk

Looks like you did it correctly to me. Yes, big difference in performance. I can't even run a 64 buffer for more than a second or two with that test session. I am mystified.

Have you experienced the other issue I have, extremely slow response to solo playist lanes (to audition takes)?

I am a comping together a performance now and it's all working A-Ok here.

Out of curiosity, are you running your 48 gig of ram across 6 of the 8 available ram slots in your machine?

[Howardk]

Originally Posted by LDS

I am a comping together a performance now and it's all working A-Ok here.

Out of curiosity, are you running your 48 gig of ram across 6 of the 8 available ram slots in your machine?

Yes, 6x8GB in the recommended slots.

I have tried El Capitan and Sierra, but not High Sierra. I have it installing onto an SSD now. I can't see it being that, but I am "in leave no stone unturned" mode and I can't see any other stones at the moment.

. . . and much thanks for running the test session!

[Howardk]

Log of Steps/Notes

2019-01-27
Removed excess slide in SSD trays from Mac Pro.
Installed High Sierra (10.13.6) on a 240GB OWC 6G SSD (I did not have to update Mac Pro Firmware, 10.13.6 had been previously installed).
I did not attach to an iCloud Account during setup.
Downloaded and installed Pro Tools 2018-12 and HD Driver.
Test session Test for CPU Spikes 2019-01-19a (Link to test session) on a HDD in secondary optical drive SATA connector.
Did a short run of test session with 256 Buffer, changed to Eleven "Lite" Native, no Piano VI. . . No surges on meter.

Installed Air and Avid Plugin Bundles.
All Playback Engine settings in this log are with "Ignore Errors" Unchecked, "Dynamic Plugins" Checked, Cache Size Normal, unless stated otherwise further on.
Did a short run of test session at 256 buffer and no CPU surges. Solid at 13%.
Changed secondary monitor to 90 degree rotation. Still all good on the meters.
Test Run 1 (256 Buffer) - CPU Total meter steady around 13%. However 4 minutes into the test run, the CPU pinned for a second and then AAE-9171 Halt Error (I happened to catch it with my eyes).
Test Run 2 (64 Buffer) - CPU Meter 25-35% running record with test session (I have not been able to even run a 64 buffer on my last install of Sierra as of late). AAE-9171 halt after 6:53. I noticed that the time of Halt (time the recording files were written) equals the session file auto backup file time (2:06). I checked the same for Test 1, and times also equal on both the recorded audio and session auto backup files (1:56).

Performance is vastly improved with this fresh installation onto High Sierra. . . but still saw this surge that causes the AAE -9171 Halt, which seems to correspond with the Auto Save Time. . . I believe someone mentioned this finding earlier in the thread.

I have not done any of the Pro Tools optimization settings yet.

Basically at this point my system is acting much better. . . no idea why. I have done fresh minimum installs onto El Capitan and Sierra in the past, and had short-lived decent but intermittent performance, and the CPU Spikes (Total CPU meter jumping) show up eventually for no known reason. . . and the AAE-9171 Halts sneak in as well.


Disabled Pro Tools Session Auto Save.
Test Run 3 (64 Buffer) - Started Test. While test running at around 23 minutes, opened Big Counter, and Expanded Transport and placed on 90 degree right side monitor. Halted after 36 minutes.
Set Energy Saver Options to Never.
App Store - Unchecked check for updates.
Test Run 4 (64 Buffer) - Test completed (117 Min).
Test Run 5 (64 Buffer) - Re-enabled PT Auto Save, session Halted after 2:37. . . did not auto-save. . .maybe because of Halt on Screen. . . or maybe Pro Tools does not autosave while rolling. If that is the case, and the save is held until Pro Tools stops, then that would explain the correlation, and maybe it will not save with the Halt Screen up. I don't know. . .
Test Run 5b (64 Buffer) - Re-running test, no changes. I am watching the Session File Backups folder, and see that Pro Tools does not Auto backup while recording. Good to know. Session just halted at 11:24 with AAE-9171, and as soon as I "ok" the error, then it wrote the Auto save/backup.
Test Run 6 (64 Buffer) - Autosave disabled again. AAI-9171 Halt at 77:57 into the test.

I submit: Auto Back/Save does not seem to be the culprit, but it tends to look like it because Auto save seems to be suspended while the session is rolling to record and write after the halt is OK'd, so the timing can look suspicious. The success of Test Run 4 is peculiar, as os the failure of Test 1. . . the random nature of the Halt would mean there is a level of "chance" to this so it may just be that.

A 64 buffer is not really required for my purposes, but it is an indicator of the systems abilities, and also to compare to Yosemite and PT 12.8.1, which seem to be just fine with a 64 buffer. . . I will shall try 128 buffer next on this fresh install of High Sierra and PT 2018-12.

Auto Back/Save re-enabled
Test Run 7 (128 Buffer) - running between 14-29% on Total CPU meter. Test completed full run, no Halt.
[Test Run 7b (same as above) - running between 14-27% on Total CPU meter. Test completed full run, no Halt.
[Test Run 8 (same as above) - I extended the test time to 470 minutes by duplicating the midi and audio content in the timeline (CMD-D). Ran over night and test completed full extended run (470 min), no Halt.

The Total CPU performance results of this install is currently more stable (less visible peaks), and the AAE-9171 Errors/Halts much reduced, compared to what I have been experiencing on El Capitan or Sierra with Pro Tools 2018-12 on this same computer/HDX Card.

I am going to CCC this High Sierra Test Volume as a baseline and contemplate my next move.

2019-01-28

Installed CCC
Changed Audio Outputs so I can listen to YouTube vids.
used to duplicate CCC Install SSD drive to a HDD (noticed it did not use an afs container on the target).
Test 9 (128 Buffer) on HDD CCC of - Halted at 5:24
Test 9b (as above) - Complete to end (117min) without a Halt.

Changed Operation Preference settings to "Automatically Create New Playlist When loop Recording".

Test 10 (128 Buffer) - Loop recorded 4 bar section to build playlist takes. Got an AAE-9171 Error/Halt after about 3 minutes.
Changed the recording tracks from Waveform to Playlist. Soloing a Playlist lane often gets colour swirly beachball and takes 1-10 seconds to actuate solo. . . this is with 5 Tracks grouped (1 second if playlist lane is close to the top, 10 seconds if toward the bottom). 1-4 seconds if the tracks are not grouped. There are about 50 lanes in these tracks and I often get to 100-200 in a normal project. This is not useable.

I switched back to my SSD OS installation, same issue. I noticed that Pro Tools is at 93% in the activity monitor, and it shoots up to 180% when I click solo on a lane (this is at rest and in the real world I am often in loop playback while auditioning playlist segments).

Changed to Yosemite PT 12.8.1 system drive. Opened same session. Pro Tools is at 80% at rest as displayed inn the Apple Activity monitor. Soloing Playlist lanes is instant and CPU usage does not jump up. Feels like good old Pro Tools 10 on Mountain Lion. . .

Loop recorded more with Yosemite/PT 12.8.1 to about 140 playlist lanes in the these tracks. Still pretty much instant Solo response with auditioning between playlist lanes, even while playing back.

Opened same session (with the 140 playback lanes) in High Sierra/PT2018-12. Clicked on a playlist solo lane when in playback. Playback halted in a second and swirly beachball for about a minute, then the AAE-1971 message popped up. CPU on Apple's Activity monitor was at 186% during this. Tried it again in loop playback. . . let it loop a few times, then soloed a playlist lane toward the bottom, this time it keep looping. . . but I had the beachball for a minute or so before I regained control. Turned the track group off. Click on 1st lane, solo engaged quickly. Clicked solo on 40th lane, beachball and about a 5 second lag before solo lit-up. 11 seconds for 90th lane, 13 secs for 140th/last lane.

Switched to my current Production System disk (Sierra/PT2018-12), which is acting up as well. Lag for Playlist lane solo lane is the same. . . long lag time.
Deleted folder /Users/howard/Library/Preferences/Avid/Pro Tools
Deleted WaveCache.wfm file in session folder
Still the same total lack of responsiveness to Playlist lane solo.
Changed to 1024 buffer. Still same issue.

This is not useable. Just me?

Still testing production computer, to see if i can reduce the AAE-9171 comparable to the minimal install.

Test 11 (128 buffer, same test session, my Production System Drive Sierra/PT2018-12)[/U][/B] - AAE-9171 Halt after 2:43 mins.
Test 11b (same as above) - AAE-9171 Halt after :52 seconds
Test 11c (same as above) - AAE-9171 Halt after 4:38
Test 11d (same as above) - AAE-9171 Halt after 26:31
Test 12 (256 Buffer) - AAE-9171 Halt after 20:05

Removed all Plugins from /Library/Application Support/Avid/Audio/Plug-Ins Folder.
Deleted "/Users/howard/Library/Preferences/Avid/Pro Tools" Folder
Shut Down
Removed Sonnet Tempo Card, and put Sample and System SSDs into tray slots.
Restart. PRAM Reset,
Reinstalled Pro_Tools_2018_12_Mac_105234_
Reinstalled HD_Driver_2018_12_Mac_105261_
Removed Syncovery Background scheduler from startup items.
Re-Installed First AIR Instruments (not content).
Opened Pro Tools and then Quit/Closed.
Installed Avid_Complete_Plug-In_Bundle_18_10_Mac_105368_

Test 13a (64 Buffer) - AAE-9171 Halt after 20:05
Test 13b (as above) - AAE-9171 Halt after less than 1 sec
Test 13c (as above) - AAE-9171 Halt after 1.5 sec
Test 13d (as above) - AAE-9171 Halt after 1 sec
Test 13e (as above. but with Edit Window Closed) - Halt after 5:11 min
Did a bunch more tests with Edit Window Open/Closed different zooms. . . noticed that it as an instant Halt when zoomed close in on the timeline, and ran a longer when zoomed out, even longer with Edit Window Closed. Usually halted when auto-scrolling window shifted, or when i manually zoomed in/out.

Test 14 (128 Buffer) - AAE-9171 Halt at 6:20
Test 14b (128 Buffer) - AAE-9171 Halt at 4:25
Test 14c (128 Buffer) - Test completed cycle (117min)

Put Sonnet Temp Card back in with Samples and Session SSD, but test session is still on HDD is optical bay.

Test 15 (128 Buffer) - 17 secs
Test 15b (as above) - 3:24
Test 15c (as above) - Halt at 7:03
Zoomed out to show the 2 hr timeline in the edit window
Test 15d (as above) - Test completed cycle (117min)

Back to High Sierra/PT 2018-12 minimum install test drive.

Zoom out to show 2hr Time Lin
Test 16 (64 Buffer) - AAE-9171 Halt after 48 seconds

[LDS]

Originally Posted by Howardk

Yes, 6x8GB in the recommended slots.

I have tried El Capitan and Sierra, but not High Sierra. I have it installing onto an SSD now. I can't see it being that, but I am "in leave no stone unturned" mode and I can't see any other stones at the moment.

. . . and much thanks for running the test session!

Contrary to the advice on the internet, I am running the full 8x8GB in my system. I tried 6x8GB - it boosted my Geekbench scores by 3000 points, but there was a very noticeable decrease in system responsiveness and smoothness so I went straight back to 8x8GB. If you have another 2x8GB of ram on hand, that might be another stone worth turning.

[the.engineer]

Originally Posted by Howardk

Unfortunately it looks like Spectrasonics current versions and Yosemite are a no go (link to compatibility info). I use Omnisphere 2.5, Keyscape, Trilian and Stylus. . . which means no PT/OS X version to land on that does not have Pro Tools CPU Spikes, and is a Spectrasonics compatible OS X version. I could install older versions of these VIs, but then I will likely not be able to go back and forth between PT/OS versions as I was hoping. . .

Current version of Native Instruments Komplete 12 is not compatible with Yosemite (Native Access will not work) and the compatible version of Native Access (1.51) is crashing when I try to set the library folder location or link to some libraries. . . I sent a post to NI.

Serum Installer did not install the Serum Presets. . . had to drag those over from another system drive that already had Serum installed, but at least that is now working.

Problems 3 for 3. . . it's a bit of a mess so far.

This is a real bummer as native instruments is a part of my usual workflow. Thanks again Howard for your continued diligence.


Sent from my iPhone using Tapatalk

[Howardk]

I wonder if there are any adverse affects from the Mac Pro (2010/2012) firmware upgrade that happens when High Sierra is installed?

I had tested High Sierra about 10 months ago, so I believe my Mac Pros have the newer firmware from that OS X upgrade I installed on a test drive.

This is some of the system information for one of these computers, found in the Hardware Overview(Click on Apple, System Report):

Model Name: Mac Pro
Model Identifier: MacPro5,1
Processor Name: 6-Core Intel Xeon
Processor Speed: 2.93 GHz
Number of Processors: 2
Total Number of Cores: 12
L2 Cache (per Core): 256 KB
L3 Cache (per Processor): 12 MB
Memory: 48 GB
Boot ROM Version: MP51.0089.B00
SMC Version (system): 1.39f11
SMC Version (processor tray): 1.39f11

I was still running Pro a Tools 10 on Mountain Lion at the time, and all seemed fine after that... I am just mentioning this because of the notion there may be problems with this firmware update that are exhibited with Pro Tools on Sierra and El Capitan which I installed months later. . . there is no substance to this at this time, it is just a thought. Any comments?

[the.engineer]

I wonder if the firmware update is to do with the Metal API.

From the Apple website:

"Metal 2 provides near-direct access to the graphics processing unit (GPU), enabling you to maximize the graphics and compute potential of your apps on iOS, macOS, and tvOS. Building on an efficient low-overhead architecture with precompiled shaders, fine-grained resource control, and multithreading support, Metal 2 evolves to give the GPU even greater control of its graphics pipeline, accelerate neural network training, and provide powerful new tools that give deep insight into your shader code."