Hi John.

Lets take the Tyros first....
Basically the Tyros loads the data at around 570KB/s the first time, which means it takes upwards of half and hour to load your sounds. That's half an hour you have to sit there loading sounds one by one to create that 1GB bank. After that point then everything is instantly available, minus the 1 minute it adds to to the bootup time. So, not all that bad other than the fact that if you want to unload, replace or add a new sound, it's going to be very slow at doing it.

Now to the Mediastation,.
Streaming is a very different animal altogether. The upper limit to how much data you can load is very much so unknown because the system does not have to load the entire sound to RAM ever. Look at it this way, let say I load a 4GB Piano sound, that does not mean 4GB of memory is needed. Only a few MB from each sample that makes up the entire sound will be loaded to RAM. This allows the RAM to provide the sound instantly when you press the key, and give enough time for the hard disk to find the rest of the sample and stream that data just in time as you play in realtime.

So, just how much data you can load is very much so unknown but you can be sure it's many many times the 1GB limit the Tyros has. I have 4GB of RAM in my mediastation which is shared memory between VSTi's, the OS and everything else. That 4GB of RAM also serves my 15GB sample library I use. So as you can see, it's not really possible to work out what the upper limit is.

Loading times for adding new sounds and unloading sounds is likley to only take milliseconds.

I uploaded a video to YouTube a few months ago you should check out. You really get to see how fas the Mediastation is at loading sounds.
http://www.youtube.com/watch?v=FwqnVgNO4bQ

Regards
James.