In most other DAWs, you send to FX busses to help reduce the CPU load, instead of loading a bunch of plugins on separate tracks. What I’m finding in Cubase appears to be the opposite. I find a higher CPU load sending multiple tracks to an FX bus versus simply adding the effect plugin as an insert to the desired tracks.
This seems so counter intuitive. This isn’t exactly a deal breaker, more of an annoyance. Is there anyone here with the expertise to answer this question? I haven’t looked at my CPU core performance, but it almost seems like it’s not spreading the work evenly between the cores when using an FX bus.
I’m using Cubase Pro 12, AMD 5900x CPU and my OS is Windows 11.
Same project, all channels’ inserts removed, new FX track created with 2C Audio Aether inserted at insert slot 1, all tracks send 1 enabled and routed to the created FX track:
.
.
.
.
Here’s is the same exact project. I removed the instances of IK Tape 24 from the busses and inserted it on the individual tracks and was able to get 16 instances running without overloading. Playback is as smooth as butter. Here’s a screen shot of that.
Your CPU load suggests the opposite of what you’re writing.
I can see 15% when used on buses and 36% on tracks.
More, the busses are stereo so use the 2-channel version of the plugin while your tracks are mono so the single channel version of the plugin (half the load).
Not sure what more is going on within the project to compare fairly but in a simple comparison test like the one I posted in a new project you can see the difference between routing multiple channels to an FX track vs inserting directly.
But I believe that the problem in your project is probably the signal flow. i.e. a track feeding a group and then another group all with inserts this pushes the multithreading side of Cubase which leads to a higher load.
Yes, you are correct, the overall CPU load is less for the bus scenario. What is interesting is the bus scenario is causing an overload at just 15% overall CPU load, while the insert scenario is playing smoothly at 36% overall CPU load. This is why I stated above that it appears to be slamming a few of my CPU cores and not evenly distributing the workload.
Yes, true, but if you think about it, I was able to insert 16 instances on mono tracks without issue. Yet, only 5 instances on stereo tracks, so roughly equal to 10 mono tracks worth of processing and it overloaded.
Later today, I’ll set up a more cut and dry example. I was just surprised by what I was finding.
so i am adding my edit on above post as a reply now cause I saw that you had replied by the time i finished editing…
So
this is not a problem with your project to be fair, it is the way the daw works.
when routing gets complicated and involves plugins then each plugin process should finish before the next one starts, say of the group that follows, etc, so this way the program has to multithread for using serial processing rather than parallel
So maybe what you’re saying about Cubase could have a point but I can not tell as I haven’t run a direct comparison with the same tracks, plugins, etc between Cubase and another daw.
That big, red square is an overload indication similar to the ones on each channel meter in the Mix Console. It indicates that something caused an audio dropout and will stay red until you manually reset it by clicking it.
In my experience this tends to come on when I change out plugins or make a change in the routing. In the context of this topic, I don’t think it holds much value. The Peak meter below is more telling about the load on the audio engine.
I think this is more of a positive side effect than anything. The Mixer in most DAWs reproduces the signal flow and workflow of large format mix consoles from the analog days where having 30 individual hardware reverb units isn’t feasible. Besides, in many cases you want to use the same reverb on multiple sources so a send bus makes sense from several perspectives.
I just want to be clear, it was more than just the indicator coming on, playback was starting and stopping, throwing overload errors that required an acknowledgement by click “OK”.
Yup, that’s exactly what I’d rely on to tell me there’s an issue rather than the indicator.
If you do a search here you’ll find several threads discussing the nuances of how cores get used & maxed out and the best ways to manage that. Often the discussion centers around VSTi’s - multitimbral vs. monotimbral for CPU useage. But the ideas are the same.
