The thing is, Bill, CPU load depends on the number of voices played. It is dynamic. Unlike hardware, where the architecture is designed to give you a certain amount of polyphony, and no more (but no less!).

So how, for instance, does the OAS determine whether you can load a VSTi or not? You might only need two voices of a streaming bass voice, but the computer certainly won't know that unless you tell it in advance.

It may indeed only let you load 4 VSTi at a time, but I guarantee you, some modern VSTi can bring a fast CPU to it's knees all by themselves... This is the achilles heel of the 'open' systems, which leverage CPU's to do the work of hardware, but do NOT guarantee you how many voices you can use, nor give you a soft landing if you exceed them...

Have you ever managed to overload your CPU on your Abacus, Bill? What happens? Does the entire arranger come to it's knees, or just the VSTi? B4 is NOT going to tax this. It's pretty CPU efficient, and uses no streaming samples (or any samples at all). This is going to occur when using layered up Performances from some of the big streaming sample libraries (Goliath, Ivory, VSL, etc.).

The whole dynamic nature of VSTi's must be a nightmare for the OAS to figure out what and how much any particular VSTi is going to effect polyphony. Once you start to use these libraries for the styles, as well as your 'live' playing, this is a brick wall you are rushing towards!

Are there any printed specs on how well the OAS handles each particular VSTi?