So, how many Pianoteq3 voices can you get in realtime from the MS?

I realize that English is not your first language, Dom, but perhaps the "(and yes, I know a four element voice uses four voices!)" sentence might have clued you to the fact that I know the difference between the total number of oscillators (or 'voices', as they are usually called, in English) and the total number of notes available for any particular sound, But feel free to keep belittling me. I can take it better than you can take legitimate questions about the MS, anyway!

Different VSTi's have vastly different CPU requirements. You may easily get 200 GIGA voices, or other sample playback systems with minimal processing on modern CPU's. But modeling is one of those things that, at ultra low latencies, can quickly drag even the beefiest CPU to it's knees. Or am I talking without any know how again?

But let me get this straight, because it's quite an achievement... If you are playing a style, or streaming some audio, and max out the CPU with something like Pianoteq3, the style still keeps on playing and the audio streaming, and only the Pianoteq3 chokes? That's pretty darn good...