Really depends on what else is going on in Max and Live. If you have other intensive processes there, there's a good chance a standalone will yield best performance. That way your Jitter rendering system is totally independent from all the rest and can't be held back by shared timing/scheduler issues and the like. And your OS will assign the process to whatever free CPU cycles it has, circumventing Max's limitations in that department.

I split up my large motion tracking / Jitter rendering / audio synthesis systems like that. I'm getting 50% fps gains from it. Of course you need to work out a workable system for sending control data back and forth, using audio signals, udpsend/receive and/or jit.net.send/recv.

Another benefit when using the networking objects is that, when needed, you can easily move a component to another computer.

hit dtr, you answered. I know it can sound obvious, but I like to have some experience and opinions from other users and especially expert like you. here, I need to analyze audio + use midi data to trigger/alter my visuals.

udpsend/receive seems to be okay here. I would analyze audio + make the visuals generations in the same standalone.

I didn't test the benefits of jit.net.X vs udpX