[BUG] CPU usage increase while idle, due to noise gate

[PeteNy]

something strange with this new 0.7.3 version an how it is using CPU for audio processing. CPU usage (for audio processing) in GP3 climbs up to 100% while not playing. once i start playing the guitar the CPU usage goes back to ~30-40%; which is still high. Although I use a busy project, 3 instances of NAM.

probably the built in Noise gate causing this.

GigPerformer3; Win10, i7-11th gen, 2x16GB RAM. Two-Notes Wall of Sound as IR loader. (Nam's IR loader is switched off)

To test this further I created a separate project. There are no other plugins active other than NAM and WoS as IR loader. Test:

1. 1x NAM:

all seems fine. Idle CPU: ~9% while playing CPU: ~9%

2. 2x NAMs:

1x pedal capture into another NAM with an amp capture: Idle CPU: ~48-50%. while playing CPU: 13%

3: 2x NAMs:

1x amp capture in each, each on a different signal (L/R). Idle CPU: ~16% while playing CPU: ~16%

4. 3x NAMs (on 2 signal paths):

1.signal path: 1x NAM with pedal capture, into another NAM with amp capture. 2.signal path: 1x NAM with an amp capture. Idle CPU: ~54-57% while playing CPU: ~20% i started to get a suspition it is due to feeding one NAM into another.

5. same as #4 but use the built-in gate on only the first NAM per channel (L/R)

Idle CPU: ~16% while playing CPU: ~16%

This was not the case with the previous version.

[sdatkinson]

Hmm...it doesn't make sense that the noise gate would be responsible.

Does this replicate for you in a DAW like Reaper? (I don't own Gig Performer.)

I set up a track in Reaper with two NAMs. I also took an input file and copied it a few times with a gap ("idle"?)

Playing the track, I see roughly equivalent CPU usage during both parts:

Silent input:

Non-silent:

While not playing back ("idle"?), the CPU usage is a little lower perhaps, but basically the same:

Lmk if you can provide more detail so I can understand what needs to be done to replicate the issue.

[PeteNy]

I will run more tests and will investigate it further. Sent from my iPhoneOn 23 May 2023, at 05:32, Steven Atkinson @.***> wrote: Hmm...it doesn't make sense that the noise gate would be responsible. Does this replicate for you in a DAW like Reaper? (I don't own Gig Performer.) I set up a track in Reaper with two NAMs. I also took an input file and copied it a few times with a gap ("idle"?)

Playing the track, I see roughly equivalent CPU usage during both parts: Silent input:

Non-silent:

While not playing back ("idle"?), the CPU usage is a little lower perhaps, but basically the same:

Lmk if you can provide more detail so I can understand what needs to be done to replicate the issue.

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: @.***>

[PeteNy]

I have run a few test.

Summary: At the end it seems that the GATE's threshold setting weirdly affects the Two-Notes Audio Wall of Sound as IR loader for some reason. Tested it with another IR loader and all was fine.

Details of my tests: GigPerformer shows the CPU usage for 'Audio Processing', while Reaper shows the overall CPU usage of computer. This is very different from each other.

I was running my Task Maanger to moitor CPU usage along with GigPerformer's CPU usage for 'Audio processing'. the below are the CPU usage for Audio Processing in Gig Performer Task Manager and Reaper showed constant, stady usage of CPU, regardless of idle or signal being processed live (playing the guitar).

Test environment: 2 instances of NAM, 1 loaded with a pedal capture and fed into another NAM with an amp capture. Two-notes Wall of Sound as IR loader after NAM (Nam's IR section were turned off on both instances of NAM).

GigPerformer 3: Test #1 - no GATEs activated: CPU (for Audio Processing) usage idle: 13% active: 13% All good.

Test #2 - both 2 GATEs are active, and on their default setting: CPU (for Audio Processing) usage idle: 13% active: 13% identical, no issues

Issue #1 idle CPU start to increase once both GATE's Threshold was set to around -66dB from here, the further the Thresholds weres adjusted the worse the CPU (for audio processing) gets.

Issue #2 once the Threshold setting on both instances of NAM reaches around -58dB, the CPU usage starts to lower again back to normal, very rapidly.

Issue #3 depending on if i route only the Left/mono output of the last NAM into the Left input of Two-notes Wall of Sound, or both Left and Right outputs of NAM into the Left and Right inputs of Two-notes Wall of Sound, the percentage differes a lot.

Threshold (-dB) idle CPU [%] to mono to stereo -80 13 13 -66 16 15 -65 20 16 -64 25 18 -63 39 25 -62 63 37 -61 75 43 -60 80 47 -59 79 46 -58 77 45 -57 73 43 -56 13 13 Test #3: once i had the above statistics, i set the Threshold of first instance of NAM's GATE to the peak CPU usage, -60dB. Threshold of the second instance of NAM's GATE is at -80dB, Defualt. Idle CPU is at 13% all good.

Once i started to increase the Threshold of the second instance of NAM, the results were similar. Idle CPU started to increase at -66dB. Peaked at -60dB, then idle CPU usage started to get lower.

Summary: Again, the above occured with Two-notes audio's Wall of Sound IR loader. I tested it with NadIR, and CPU usage was all as expected.

I don't understand why and how NAM's Gate settings would trigger this weird CPU usage incrase in GigPerformer3, idle only, and with Wall of Sound.

Reaper: everything works as should, (apart from the known issue of one side is louder) regardless of how many instances i run with GATE active and/or their Threshold values.

Let me know if i could do any more tests. I work with Two-notes Audio (in a closed group selected by their QA lead) as a beta QA tester for their new plugin. I can't talk about the product, really, however, i could perhaps do some tests on NAM if you have any specific requests. Or, i could ask their QA team to look into this matter if you want.

Let me know. Thanks for your time.

On Tue, 23 May 2023 at 07:16, Nyeste Peter @.***> wrote:

I will run more tests and will investigate it further.

Sent from my iPhone

On 23 May 2023, at 05:32, Steven Atkinson @.***> wrote:



Hmm...it doesn't make sense that the noise gate would be responsible.

Does this replicate for you in a DAW like Reaper? (I don't own Gig Performer.)

I set up a track in Reaper with two NAMs. I also took an input file and copied it a few times with a gap ("idle"?)

[image: image] https://user-images.githubusercontent.com/12240186/240130070-fb454b51-a3fe-47b6-b8f7-6f14b89e89cb.png

Playing the track, I see roughly equivalent CPU usage during both parts:

Silent input:

[image: image] https://user-images.githubusercontent.com/12240186/240130321-98054b19-702e-406b-a8fe-cee247dfed2b.png

Non-silent:

[image: image] https://user-images.githubusercontent.com/12240186/240130209-12094175-8743-45da-ac6e-ed0ad4daee5f.png

While not playing back ("idle"?), the CPU usage is a little lower perhaps, but basically the same:

[image: image] https://user-images.githubusercontent.com/12240186/240130537-57cf80d8-ad39-4fd5-9f5d-775680a6cd3d.png

Lmk if you can provide more detail so I can understand what needs to be done to replicate the issue.

— Reply to this email directly, view it on GitHub https://github.com/sdatkinson/NeuralAmpModelerPlugin/issues/255#issuecomment-1558504938, or unsubscribe https://github.com/notifications/unsubscribe-auth/A72SPZUYAQ6B5M7ZHSBVT6DXHQ4WPANCNFSM6AAAAAAYAO35UA . You are receiving this because you authored the thread.Message ID: @.***>