Hmm... I think Tobias has a point. Throughput is the limiting factor in 
many Csound instruments (as it is on many video manipulation programs).
Watch what happens to compile times when you add part of the instrument
that writes and reads memory constantly (ie: reverb). To double the size
of these writes and fetches will probably slow csound down a great deal.
As I understand it, SGI's O2 has spent a great deal of time on this
bus speed problem, and the remarkable speed up in real-time that
has been reported here is more due to this improved io, not the
speed of the processor. My friend in scientific visualisation
at the FloridaState SuperComputerResearchInstitute *swears* that
this new buss design is the cat's pyjamas (and boy, he knows real-time
video as well as anyone I've met!)

Well, I'm ready tobe corrected, but I feel that processor speed is only
one of the issues in speed of csound execution...


CharlieB