|
|
|
12-22-2022, 06:31 PM
|
#1
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
[6.72] Bad performance with multichannel plugins
After installing 6.72, the project I'm currently working on started to put a very heavy load on the CPU (fans went full throttle) and audio was just a bunch of glitches.
Same session reopened with 6.71 worked like a charm.
After much investigation, I found that the plugins that drew so much were the numerous instances of Fabfilter Pro-L2 I have on atmos (10 channels) tracks.
I also have quite a few instances of Penteo 16 pro, Cinematics Room, ReaSurround and various SoundParticles plugins (all 10 channels), but the FabFilter seem to be the only one concerned with 6.72 but not at all with 6.71.
It's hard for me to tell if it's a Reaper or FabFilter issue, but the fact it works great with 6.71 suggests me that I should start by posting here.
All the best,
p@T
Mac Trashcan (Intel) / OS 12.6.2 (Monterey)
Both Universal and x86_64 builds have the same behaviour
|
|
|
12-23-2022, 06:31 AM
|
#2
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
How does 6.72 compare against 6.70?
If you look in the performance meter, what does that track FX CPU readout show for each version for that track?
The reason I ask is 6.71 had some stuff in it to report very granular silence information to VST3s, however a lot of plug-ins fail miserably when they get this information, so 6.72 removed it (but 6.70 and 6.72 should behave similarly, hopefully! if they don't, then I need to look elsewhere...).
Last edited by Justin; 12-23-2022 at 06:50 AM.
|
|
|
12-23-2022, 09:04 AM
|
#3
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
How does 6.72 compare against 6.70?
|
Spot on, 6.70 & 6.72 severely lag with around 40% RT CPU and tons of RT xruns when 6.71 quietly sits at 2.5% RT CPU and no RT xruns...
|
|
|
12-23-2022, 01:07 PM
|
#4
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
How does 6.72 compare against 6.70?
If you look in the performance meter, what does that track FX CPU readout show for each version for that track?
|
Sorry, I misread your message and gave the overall CPU charge.
For individual tracks, especially those with FabFilter, 6.70/72 are at around 1% when they are at 0.04% with 6.71.
My "peak cpu" track has 3 10 channels plugins (Flux Epure, Flux Solera and Penteo16 pro) is at 2.45% with 6.70/72 and 1.06% with 6.71
Hope this helps :-)
|
|
|
12-23-2022, 03:17 PM
|
#5
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Quote:
Originally Posted by matnoir
Spot on, 6.70 & 6.72 severely lag with around 40% RT CPU and tons of RT xruns when 6.71 quietly sits at 2.5% RT CPU and no RT xruns...
|
If you tweak the block size or anticipative FX so that 6.72 doesn't RT xrun, do they sound the same?
Do many of those 10 channels on that one highest-cpu track have silence most of the time?
|
|
|
12-25-2022, 05:37 AM
|
#6
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
If you tweak the block size or anticipative FX so that 6.72 doesn't RT xrun, do they sound the same?
|
My block size already was at 3072 (I usually am around 512/1024, but I did increase for this relatively complicated mix), so I did not try to change it as I do believe, possibly wrongly, but I recall reading that somewhere once, that above 2048 does not make a real difference).
The anticipative FX, even at 1200 does not completely suppresses xruns. (however, I don't know if that may help, 6.73 was kinda ok at 800ms.) But, in both cases, the "peak cpu track" now goes up to almost 4%.
Quote:
Originally Posted by Justin
Do many of those 10 channels on that one highest-cpu track have silence most of the time?
|
Yes, they do :-)
Merry Christmas to you and your dear ones,
p@T
|
|
|
12-25-2022, 07:16 AM
|
#7
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
For high channel counts you might find better performance with lower block sizes, eg 256, due to cache effects. Anticipative FX at 1000ms with that should be plenty.
|
|
|
12-25-2022, 02:13 PM
|
#8
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
For high channel counts you might find better performance with lower block sizes, eg 256, due to cache effects. Anticipative FX at 1000ms with that should be plenty.
|
Thanks for the advice. 256 is definitely not an option, but 6.71 handles 512 well (unlike 6.72).
The anticipative FX setting had a side effect I had not noticed before. I'm using the Dolby Atmos Music Panner to move objects around. It does not use audio at all, it's merely some kind of remote, indicating to the Dolby Renderer the object position. When the anticipative FX is set beyond 460ms, the object metadata seems no longer sent to the Dolby Renderer, and the objects no longer follows the automation and stays at a fixed position. I tried to play with the ignore tail setting, but to no avail. So I ended up setting the anticipative FX to 350ms (as I feared a larger value might noticeably shift the metadata in time)
|
|
|
12-26-2022, 07:44 AM
|
#9
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Adding a compatibility setting so you can enable this for plug-ins that support it to operate more efficiently!
|
|
|
12-27-2022, 05:21 PM
|
#10
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
The latest +dev builds have a compatibility setting to re-enable this for each plug-in, if you want to try!
|
|
|
01-04-2023, 01:45 PM
|
#11
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
The latest +dev builds have a compatibility setting to re-enable this for each plug-in, if you want to try!
|
Thanks Justin and a vey happy new year to you and your dear ones.
Just tried the dev build (673+dev0104), but it's still the same. When the anticipative FX processing render ahead is set beyond 460ms, the Dolby Atmos Music Panner no longer sends automated metadata to the Dolby Rendering App...
|
|
|
01-04-2023, 02:37 PM
|
#12
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Quote:
Originally Posted by matnoir
Thanks Justin and a vey happy new year to you and your dear ones.
Just tried the dev build (673+dev0104), but it's still the same. When the anticipative FX processing render ahead is set beyond 460ms, the Dolby Atmos Music Panner no longer sends automated metadata to the Dolby Rendering App...
|
That's a separate issue - there's likely some Atmos limitation that will require you to keep anticipative FX renderahead low.
The question is: how does the performance compare to 6.71, if you enable the compatibility setting for per-channel silence notifications on those performance-sensitive 10-channel plug-in instances you mentioned.
|
|
|
01-05-2023, 01:54 PM
|
#13
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
The question is: how does the performance compare to 6.71, if you enable the compatibility setting for per-channel silence notifications on those performance-sensitive 10-channel plug-in instances you mentioned.
|
I tried all options, i.e keeping the auto-bypass on at project level but forcing it off on the problematic plugin, setting it to off at project level and forcing it on or off at track level.
I also tried to change the threshold value at project level (to -140dB s my session is in 24bits).
In all configurations the high-cpu track remained more or less at 4.5% with short spikes above 5% when the audio starts.
I checked again with 6.71 and the very same track is at 0.08%
|
|
|
01-05-2023, 03:27 PM
|
#14
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
To be clear, the per-plugin compatibility setting you need to enable is:
"Pass per-channel silence information to plug-in"
For the ones on that track -- note that this will break many plug-ins that don't properly implement this feature, so you can't enable it everywhere.
If you enable this for the plug-ins in question and keep everything else as it was in 6.71, it should behave identically to 6.71.
|
|
|
01-05-2023, 04:40 PM
|
#15
|
Human being with feelings
Join Date: May 2019
Location: France
Posts: 41
|
Quote:
Originally Posted by Justin
To be clear, the per-plugin compatibility setting you need to enable is:
"Pass per-channel silence information to plug-in"
For the ones on that track -- note that this will break many plug-ins that don't properly implement this feature, so you can't enable it everywhere.
If you enable this for the plug-ins in question and keep everything else as it was in 6.71, it should behave identically to 6.71.
|
You're a marvel! (but we both knew that already) Yes it does.
Just out of curiosity, any general rule of thumb as to which kind of plugins will break ? (I'm not asking for brand names of course...)
Keep up the good work, thanks again!
|
|
|
01-05-2023, 07:27 PM
|
#16
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Quote:
Originally Posted by matnoir
You're a marvel! (but we both knew that already) Yes it does.
Just out of curiosity, any general rule of thumb as to which kind of plugins will break ? (I'm not asking for brand names of course...)
Keep up the good work, thanks again!
|
No idea — if you’re daring you can enable it for all of them and then disable when you find some misbehave…
but in general I would enable it on plugins that you find to be particularly intensive, and especially if they process a lot of channels, then test to make sure they work right
|
|
|
Thread Tools |
|
Display Modes |
Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 06:52 AM.
|