Using instrument tracks against MIDI tracks

Hi all,

I am interested in your opinion of using instrument tracks against MIDI tracks (whatever you prefer)
You see in the attached picture „example 1“ that for each of the three instruments, Viola, Cello, Bass, I loaded a seperate instance oft he corresponding VST instrument

Another way of doing it would be to only load for the first instrument track the corresponding VST instrument and adding up to 8 (stereo) MIDI tracks which are automatically connected to the first instrument track (and the VST instrument itself)
This is shown in picture „example 2a“.

Picture „example 2b“ shows that not only the MIDI tracks are appearing in your project zone or mixer zone, it additionally shows corresponding VST tracks which are linked to the first instrument track.

To analyze it from a tecnical view:

Example 1 (one instance of VST instrument per instrument track)

  • Every instrument track can be handled das an audio track, that means you can directly put inserts/effects/EQ on it
  • less stracks in your project/mixer window = better overview
  • Needs more system resources

Example 2 (one instance of VST instrument + MIDI tracks)
• Advantage:

  • Needs less system resources
    • Disadvantage
  • increased effort to use inserts/effects/EQ as you cannot put them on the MIDI track. You have to use the corresponding VST track instead (not the VST instrument track)

How are you dealing with this topic, creating a compromise between better workflow with less tracks to handle with and using MIDI tracks for saving system ressources?


I use MIDI tracks for multitimbral VSTs like Halion Sonic and Omnisphere. Otherwise I use Instrument tracks.

You don’t necessarily have a negative effect using more system resources, mainly RAM. Instruments loaded on individual tracks use more threads, and therefore spread the CPU load better than using one instrument multitimbral.

Yeah what peakae said.

The OP might want to do a search in the Cubase 9 (and 8?) forum where there are threads that cover this topic in detail.

I like to point out that CPU capacity is transient, you can’t save it. Every CPU cycle on a computer always gets consumed - always. It either gets used doing some useful work for you or it gets discarded unused. So if you are doing things to reduce CPU consumption that also makes it harder not easier for you to work that’s a bad trade-off (unless of course you are at risk of consuming all the capacity).

Personally I tend to concentrate on MIDI tracks, and rendering, convenience etc. comes at a subsequent phase. This is probably because (a) I’m old school and grew up with MIDI and (b) I like the flexibility that MIDI gives me.

I hate having to decide on the instrument at the point in time when I create the track! For example, if I start a tune with an instrument track, then decide at a later stage that I’d like to hear that on some external MIDI instrument, I have to create a new MIDI track routed to the instrument and copy the part to that new track.

On the other hand I totally get the rationale behind instrument tracks; it’s much easier for someone starting off to get to grips with – why should they have to struggle with a paradigm based on hardware that is older that they are?

Totally get that perspective. I pretty much use Instrument Tracks exclusively nowadays. But I don’t know about the old school aspect impacting that - my first midi sequencer was text based. :mrgreen:

See my avatar? :smiley:

I got the drift. But even with a magnifying glass could not determine the specifics. :open_mouth:

What was the one with the command line; and the one with the colored columns?

(this has gone way off topic!) It’s C-Lab Supertrack on a Commodore 64. The GIF shows the sequence of switching on the the C64 (which booted into BASIC) and then loading the program from the 1541 5.25" floppy disc. In real time, that all used to take about 10 minutes.

Thanks for all of your replies so far.

Yeah, I know that fact, but thank’s for giving that hint. Of course, I would load at least the amount of instrument instances as the total amount of CPU cores I have (in this case, 4). Spreading all instruments, let’s say, about 20 to 30 over 4 instances at least sounds to me a reasonable way. So in that case, I would use at least 4 multitimbral instruments. Nevertheless, I would have to deal with all that MIDI tracks…

I think I need to test the overall performance of my PC when doing it the one or the other way. Maybe there is no difference at all…who knows. I think it depends on the project size. As I am quite new to composing orchestral stuff, my projects surely will be not that big.

At this time, the only reason what keeps me from using only instrument tracks is that I can not use the instrument built in reverb module efficiently (in this case the EastWest Spaces Reverb). I would have to load it on any single track if using instument tracks. What an effort to do that…and quite confusing.
A configuration as a multitimbral instrument at least lets me set a reverb for all MIDI tracks which are connected to that instrument instance.
Or I have to use external reverb plugins…
Well, I think there is no “best” way, I have to try different things out

I always disable any “built-in” reverb that might be in the VSTi, and instead configure sends to an FX track with a reverb on it. That uses much less CPU, you can have more control over it and (particularly when trying to mimic real instruments in a real space) it results in a more coherent and believable effect.

That’s a great point. Had never considered the resource implications of reverbs in VSTi’s, only the aesthetic ones.

To the OP if you run what you figure is a typical Project for what you are currently doing (i.e. typical track counts, mix of used VSTi’s, fx, etc.) what are you seeing in the Windows Task & Performance Monitors for CPU and memory use; and also the Cubase VST meter?

I have to set up a new project with my new instrument which is capable of setting multitimbral functionality. This might take a while. I will come back and edit this post later.

For some reason, I lost focus on this topic…but now I have made the test.

21 Instruments playing in a 38 second piece

Multitimbral Set-up: 5 Instances of PLAY Engine => 5 Instrument Tracks + 21 MIDI Tracks
Single Instrument Set-up: 21 Instances of PLAY Engine => 21 Instrument Tracks

Just for making it more transparent, I have loaded up three pictures with one of them showing the instrumets and their articulation and the other one’s showing the measured performance level in the Windows Task Bar and Cubase Audio Performance.
I let the piece run in cycles for 5 times for each configuration and made screenshots of the highest load levels.

As you guys said before, there is no much difference in CPU or RAM load but I recognized that in the single instrument set-up I sometimes had hard peaks (and audo drop outs) in the Cubase Audio Performance level but they did not occur in every cycle. Don’t know why, as there are no more ressources used.
I think freezing some instruments could help here

FWIW i’ve found i get better performance using multiple VSTis rather than one multitimbral instance with mutiple midi and output channels, less peaks etc too.