Question about performance: How I bring an M1Pro to its knees with a single DrumKitDesigner.

[Sir Hannes]

 

Overloads and complete dropouts on a M1Pro 16" MacBookPro, I/O buffer 1024 samples, 200ms roundtrip (due to airpodspro), 10 threads selected (de-selected: automatically). 😞  Is DrumKitDesigner really too much for my system? In use I have not so much going on: 1x LogicSampler, 2x Strings, 3x instances Hive, 2x audio tracks in Flex mode, 1x DrumKits (Asian Kit) and that DrumKitDesigner (DKD). Sure, the masterplugs are intense, but I have sessions where 140 tracks plus the same master chain plays and the M1Pro, well, looks kind of bored in terms of CPU. Now playing this DKD in solo creates this 100% spike. But the session at once works. Cycle-Mode let the audio crackles, then it goes silent (stereo output meter still shows Logic playing). No external hardware in use at the moment. For PlugIns see picture. I'd really like to hear your opinion, as this tiny session gets annoying unstable and I wonder how I would continue finishing it.

Continue working I have to close this session. I re-open it. Hopefully the audio is back. But after 2nd or 3rd time, this failures, no audio is coming back and I have to shut down Logic and reload it. As I write this in the LogicProHelp forum, I suddenly hear the Drum Designer loop coming from the parallel universe again out of nothing as I closed the session! ("empty the buffer finally"?). The CPU window is still open and all the cores are spiking around. What do they process if there's nothing on?

If I throw out DKD, this whole mass stops, all CPU cores go down to where they are always here on my system: around 25%.

Here's the screenshot I made in playback mode. Pressing stop let the spike stay where it is. I feel really stupid right now.

 

 

 

Edited March 28 by Sir Hannes

[des99]

You're overloading the Live Core. Try to select a different, non-DKD track, so the whole thing and all it's plugins isn't all loaded onto one core.

Also - why so high a buffer? Or ARM, higher buffers are generally *less performant* than low ones. Drop it down to 64 samples, and see how you go.