Start a new topic

DAPS CPU soaring high after 12-core CPU upgrade to Mac Pro

CPU % was much lower with my 6 core.  This is just idle.  Pro Tools is not even running.

image

image



Hi Jon,


I don't know what the issue is offhand.   Just to confirm this is a 2013 MacPro with an upgrade to Xeon E5-2697v2 processor correct?   What Ram upgrades were part of this?  


I will ask if any of the team has any experience with this.   I know this isn't a configuration we have tested.


Best,

Adam

Hi, yes RAM was upgraded too, but it was done first and there was no issues with CPU spikes. I just did the 12-core install yesterday and am seeing these spikes. Looks like it could be _coreaudiod related as well so maybe it’s macOS and not you guys.. Coreaudio is over active with DAPS closed. Item: OWC 12-Core 2.7GHz Intel Xeon E5-2697 v2 Processor Upgrade Kit for Mac Pro (Late 2013) Item: 128.0GB OWC Matched Set (4x 32GB) PC3-10600 1333MHz DDR3 ECC-R SDRAM Modules for Mac Pro Late 2013. New, Lifetime Advance Replacement Limited Warranty. (OWC1333D3Z3M128)

I'm not seeing quite the same thing on my 16-core Mac Pro 7,1.  But I still see 13% on the CPU meter when PT isn't running. That's when source is Input.  When I change source over to Master, it does a few periodic jumps to 22%. Either of those numbers does sound high when there's nothing at all passing through the renderer.


FWIW, it appears that the renderer is only using the first 8 physical cores of my machine.  That means that a busy PT session can hit overload on those cores where there are still quite a few cycles available on cores 9-16.

Thanks Michael. Adam, yes this is on a Late 2013 Mac Pro 6,1

With core audio out of the picture (set DAPS to Send/Rtn plug), CPU drops to normal.  Seems it is a coreaudio issue.. 

image

image

image


That makes sense.  I had a session that was running fine (almost--but not quite out of CPU) on my 2013 Trashcan. I was running Send/Return on that one.  Same session is pretty strained on my 16-core, using the bridge.

Hmm, I am actually unsure now -- with some re-defining what's in /Library/Preferences/Audio, I was able to make CoreAudio chill out, but DAPS is still acting stressed for no reason.

image

image

image

Hi Jon,


If you disable re-render processing (or reduce the number of re-renders),  turn off binaural processing,  turn off loudness processing how far does this go down?


If you go into record what does the the Audio CPU in the Renderer go up to?   With and without the above enabled?  If it isn't going over or spiking it will still work.    I'm frankly not a fan of this meter as it mostly serves to increase anxiety rather than provide useful feedback.


We aren't limiting the total number of cores that can be used so this is still unexplained at this point.   I'll let you know if I hear from the team of any other potential reasons for the heightened activity.


Best,

Adam

If you disable re-render processing (or reduce the number of re-renders),  turn off binaural processing,  turn off loudness processing how far does this go down?


A little bit down, not too much.


I found that it was mainly a CoreAudio issue.  And the issue with that was mainly with using Loopback 2.  When I disable loopback, CoreAudio calms down and DAPS is good to go.  Did some master file record tests and everything seemed OK, humming under 15% (empty PT session, no objects, etc,).

Cool.  15% is what I'd expect.   Loopback2 as in the RogueAmoeba app?  What is the use case with DAPS if you don't mind me asking?   Were you using Loopback2 with DAPs prior to the CPU upgrade or did these happen simultaneously?


Thanks,

Adam

I'm still seeing 15-20% right after a fresh boot of the system (and I'm not using any Rogue Amoeba stuff).  With no signal going in, I'm still baffled as to how the DAB can chew up all of those cycles.  If I use Core Audio to drive my Focusrite 16Line, it barely registers.


There is one additional little gotcha to be aware of that should have nothing to do with Dolby.  After a fresh boot, the Avid App Manager Helper is burning about 30% of the system resources for several minutes.  This shows up in the Dolby Renderer CPU meter when it's actually got nothing to do with Dolby.  Perhaps that's just a reflection of what you don't like in that particular meter, Adam.  Since I'm not sure what the App Manager Helper is actually helping, I've once again uninstalled it.

Yeah I didn't want to tell you guys, because I was kind of hacking DAPS in order to print speaker outputs and re-renders from back to PT and didn't want you to patch DAPS and prevent me from doing that.  But basically I was sending DAPS to out a 64ch Source Nexus device, pulling the headphones out in Loopback2, sending that back to a virtual input in my UAD Console for "monitoring" while retaining all the other outputs (non-headphone only mode) for prints back in PT via Source Nexus.  I had another instance in Loopback to monitor all the output "meters" of DAPS also.

image

image

image

Hi Guys,  Interesting.   I don't know that we could or have any interest in preventing this?   I'm sure this does save some time vs offline re-renders and importing into your DAW but is a lot of hoops to jump through and an evident CPU hit?  Also I see you are using MTC.   This works but shouldn't be used for punch-ins.  Any reason you aren't using the LTC Generator plugin that comes with the v3.4 installer?


Re: CPU usage in idle.   The Renderer is processing even it if isn't receiving input so 15% while idling is expected.


Good tip re: the Avid helper.  


Thanks,

Adam

Great and the reason I print or bring back to PT is I have to deliver discrete PCM wavs in particular track orders and to monitor loudness.  The 9.1.6 re-render will make it so I might not need to do that anymore, but I may still need to output 9.1.4 in which case I was printing speaker outputs.  (This is usually done via RMU but on DAPS during work from home).  There's no way to "customize" re-renders and make a 9.1.4 is there?  I'd prefer to make all my stereo down-mix and splits all at once as well, so if I can customize re-renders, it would really solve all this.


Haven't messed with the LTC Generator yet, I will definitely take a look! Noticed some latency in preliminary tests last night.  Would I just need to use the new offsets?  Or do you think LTC Generator would solve for this?

image

Login to post a comment