I feel like there should be a hard-and-fast-rules sort of manual for best practices with regard to running large templates. And I'm curious what others' experiences are.
Currently I have a very large template being run from a Win7 Slave machine into a Mac machine.
- Slave = Win7 (newest update) with 48 GB RAM and 12 Core i7
- DAW = Cubase 7.0.6 (Mac 10.8.5)
- VEP = 5.2.12808
- Kontakt = 5.3
I'm running my entire template from only 3 instances of VEP within the Server. Each instance constains:
- 8 MIDI Ports OUT
- 10 Stereo Audio Ports OUT
- 4 Threads allocated per Instance. (Sincer there are 3 instances and 12 Cores available)
So, questions...
- Does the high number of Audio Outputs per Instance raise CPU and/or RAM use, and if so, how significantly?
- Does the high number of MIDI OUT Ports raise CPU and/or RAM use, and if so, how significantly?
- Does the low number of VEP Instances help save on either CPU or RAM use?
- Is it really best to turn off Multi-threading from within Kontakt if it is being utilized as plugins inside of VEP? (I ask because I've tried it both ways and witnessed no differences. Part of why it seems it might be smart to keep multi-threading on in Kontakt is that there are some Kontakt instances which may be HUGE,... such as LASS Violins, for example, which uses more than 2 GB within a single Kontakt; this seems like, if multi-threading were to be turned off, it might place a very large burden on just 1 core.)
Depending on the answers to the above...
- If additional VEP Instances do NOT add much burden to CPU or RAM, but additional Audio OUTs or MIDI Ports do, then it would make sense to have *more* VEP Instances broken up into smaller bits.
- If additional VEP Instances DO add a significant burden on CPU or RAM, but additional Audio OUTs or MIDI Ports don't, then having *fewer* (but larger) VEP Instances would be best practice.
So which is it?
Any answers are much appreaciated.