I was curious to understand what an Engine Cycle is. Is Engine Cycle the same as Time Load?
I finished creating a rack that has the following:
UVI Piano
Kontakt 1 (with 3 patches)
Kontakt 2 (with 5 patches)
Omnisphere 1 (with 3 patches)
Omnisphere 2 (with 4 patches)
No MIDI is sent to any of them until I turn on their route through bindings on my keyboard. Yet, I’m getting high Engine Cycles at times and many page faults. On rare occasions, I see spikes of engine time up to 2300%. But for the most part it seems to operate smoothly. I’m curious, though. Is this setup too complicated? I figured my laptop was powerful enough to handle this. I’m thinking the issue must be something else.
some observations looking at the profiler snapshot you shared:
your Piano V2 (Arturia?) plugin seems to be prone to some bursts of CPU load 53% peak vs. 2,7% average - have you tried alternatives like Addictive Keys (I use it as my main live piano, it’s pretty predictable regarding CPU load).
What worries me most are the big numbers for input processing - especially since you don’t process any audio. Looks like your audio interface is putting some intermittent load on your CPU, which it shouldn’t
Generally, I wouldn’t recommend the direct digital connection via Dante virtual soundcard setup - it creates interdependencies between your setup and the mixing setup, which may limit flexibility and troubleshooting.
The more robust solution is to go the traditional way and use a decent low-latency audio interface (Focusrite, Zoom, RME) and deliver analog signals to the mixer. Given the quality of current audio interfaces, signal degradation is minimal - and you’re completely independent of the mixer hardware. Also, the interface takes on some of the buffer processing work away from your laptop (for the virtual soundcard, there’s more real-time processing happening within your laptop). Maybe this could fix some of the input processing issues.
Generally, sour laptop should have more than enough horsepower for glitch-free audio processing, although Omnisphere and Kontakt can be resource hogs…
But running LatencyMon, as suggested by @Alexander, could be a good idea - it can identify driver issues outside of Cantabile that limit your laptop’s ability to fill an audio buffer in time.
I just chime in to add some snippets to the information Torsten gave:
You use DANTE Virtual Soundcard which is a nice idea at first glance. I gave it a try some months ago and have to say that the performance I got was far from what I was used to with any of my RME interfaces. Obviously the transformation process to dante puts quite some load to the cpu. In my case I had to increase buffer size from 128 to 512 samples to prohit crackle (that was the point when I dropped this experiment😉).
The fact that Audinate offers dedicated (and pricy!) hardware solutions (only for desktop pcs) also points into this direction.